diff --git a/README_ch.md b/README_ch.md index 7d29a6a37ec46fe9faaf83912a5f8680b11b5105..588af59c80a88c69dae8df97fa4b3eb9a816fc1a 100644 --- a/README_ch.md +++ b/README_ch.md @@ -4,16 +4,18 @@ PaddleOCR旨在打造一套丰富、领先、且实用的OCR工具库，助力使用者训练出更好的模型，并应用落地。 **近期更新** +- 2020.12.07 [FAQ](./doc/doc_ch/FAQ.md)新增5个高频问题，总数124个，并且计划以后每周一都会更新，欢迎大家持续关注。 +- 2020.11.25 更新半自动标注工具[PPOCRLabel](./PPOCRLabel/README.md)，辅助开发者高效完成标注任务，输出格式与PP-OCR训练任务完美衔接。 - 2020.9.22 更新PP-OCR技术文章，https://arxiv.org/abs/2009.09941 -- 2020.9.19 更新超轻量压缩ppocr_mobile_slim系列模型，整体模型3.5M(详见[PP-OCR Pipline](#PP-OCR))，适合在移动端部署使用。[模型下载](#模型下载) +- 2020.9.19 更新超轻量压缩ppocr_mobile_slim系列模型，整体模型3.5M(详见[PP-OCR Pipeline](#PP-OCR))，适合在移动端部署使用。[模型下载](#模型下载) - 2020.9.17 更新超轻量ppocr_mobile系列和通用ppocr_server系列中英文ocr模型，媲美商业效果。[模型下载](#模型下载) - 2020.9.17 更新[英文识别模型](./doc/doc_ch/models_list.md#英文识别模型)和[多语言识别模型](doc/doc_ch/models_list.md#多语言识别模型)，已支持`德语、法语、日语、韩语`，更多语种识别模型将持续更新。 -- 2020.8.26 更新OCR相关的84个常见问题及解答，具体参考[FAQ](./doc/doc_ch/FAQ.md) - 2020.8.24 支持通过whl包安装使用PaddleOCR，具体参考[Paddleocr Package使用说明](./doc/doc_ch/whl.md) - 2020.8.21 更新8月18日B站直播课回放和PPT，课节2，易学易用的OCR工具大礼包，[获取地址](https://aistudio.baidu.com/aistudio/education/group/info/1519) - [More](./doc/doc_ch/update.md) + ## 特性 - PPOCR系列高质量预训练模型，准确的识别效果 @@ -48,15 +50,14 @@ PaddleOCR旨在打造一套丰富、领先、且实用的OCR工具库，助力 - 代码体验：从[快速安装](./doc/doc_ch/installation.md) 开始 -## PP-OCR 1.1系列模型列表（9月17日更新） +## PP-OCR 2.0系列模型列表（更新中） | 模型简介 | 模型名称 |推荐场景 | 检测模型 | 方向分类器 | 识别模型 | | ------------ | --------------- | ----------------|---- | ---------- | -------- | -| 中英文超轻量OCR模型（8.1M） | ch_ppocr_mobile_v1.1_xx |移动端&服务器端|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar)|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_pre.tar) | -| 中英文通用OCR模型（155.1M） |ch_ppocr_server_v1.1_xx|服务器端 |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_pre.tar) | -| 中英文超轻量压缩OCR模型（3.5M） | ch_ppocr_mobile_slim_v1.1_xx| 移动端 |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/det/ch_ppocr_mobile_v1.1_det_prune_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_det_prune_opt.nb) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_quant_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_cls_quant_opt.nb)| [推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/rec/ch_ppocr_mobile_v1.1_rec_quant_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_rec_quant_opt.nb)| +| 中英文超轻量OCR模型（8.1M） | ch_ppocr_mobile_v2.0_xx |移动端&服务器端|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +| 中英文通用OCR模型（143M） |ch_ppocr_server_v2.0_xx|服务器端 |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | -更多模型下载（包括多语言），可以参考[PP-OCR v1.1 系列模型下载](./doc/doc_ch/models_list.md) +更多模型下载（包括多语言），可以参考[PP-OCR v2.0 系列模型下载](./doc/doc_ch/models_list.md) ## 文档教程 - [快速安装](./doc/doc_ch/installation.md) @@ -141,6 +142,7 @@ PP-OCR是一个实用的超轻量OCR系统。主要由DB文本检测、检测框 ## 贡献代码 我们非常欢迎你为PaddleOCR贡献代码，也十分感谢你的反馈。 + - 非常感谢 [Khanh Tran](https://github.com/xxxpsyduck) 和 [Karl Horky](https://github.com/karlhorky) 贡献修改英文文档 - 非常感谢 [zhangxin](https://github.com/ZhangXinNan)([Blog](https://blog.csdn.net/sdlypyzq)) 贡献新的可视化方式、添加.gitgnore、处理手动设置PYTHONPATH环境变量的问题 - 非常感谢 [lyl120117](https://github.com/lyl120117) 贡献打印网络结构的代码 @@ -148,3 +150,6 @@ PP-OCR是一个实用的超轻量OCR系统。主要由DB文本检测、检测框 - 非常感谢 [authorfu](https://github.com/authorfu) 贡献Android和[xiadeye](https://github.com/xiadeye) 贡献IOS的demo代码 - 非常感谢 [BeyondYourself](https://github.com/BeyondYourself) 给PaddleOCR提了很多非常棒的建议，并简化了PaddleOCR的部分代码风格。 - 非常感谢 [tangmq](https://gitee.com/tangmq) 给PaddleOCR增加Docker化部署服务，支持快速发布可调用的Restful API服务。 +- 非常感谢 [lijinhan](https://github.com/lijinhan) 给PaddleOCR增加java SpringBoot 调用OCR Hubserving接口完成对OCR服务化部署的使用。 +- 非常感谢 [Mejans](https://github.com/Mejans) 给PaddleOCR增加新语言奥克西坦语Occitan的字典和语料。 +- 非常感谢 [Evezerest](https://github.com/Evezerest)， [ninetailskim](https://github.com/ninetailskim)， [edencfc](https://github.com/edencfc)， [BeyondYourself](https://github.com/BeyondYourself)， [1084667371](https://github.com/1084667371) 贡献了PPOCRLabel的完整代码。 diff --git a/README_en.md b/README_en.md index 37250da2cd3f6ccee76b522bf10745ecb8cd649e..9e839c448101403ec35b4f1fa58cef46ecb045bc 100644 --- a/README_en.md +++ b/README_en.md @@ -1,32 +1,48 @@ -English | [简体中文](README.md) +English | [简体中文](README_ch.md) ## Introduction -PaddleOCR aims to create rich, leading, and practical OCR tools that help users train better models and apply them into practice. +PaddleOCR aims to create multilingual, awesome, leading, and practical OCR tools that help users train better models and apply them into practice. **Recent updates** -- 2020.8.24 Support the use of PaddleOCR through whl package installation，pelease refer [PaddleOCR Package](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_en/whl_en.md) -- 2020.8.16, Release text detection algorithm [SAST](https://arxiv.org/abs/1908.05498) and text recognition algorithm [SRN](https://arxiv.org/abs/2003.12294) -- 2020.7.23, Release the playback and PPT of live class on BiliBili station, PaddleOCR Introduction, [address](https://aistudio.baidu.com/aistudio/course/introduce/1519) -- 2020.7.15, Add mobile App demo , support both iOS and Android ( based on easyedge and Paddle Lite) -- 2020.7.15, Improve the deployment ability, add the C + + inference , serving deployment. In addition, the benchmarks of the ultra-lightweight OCR model are provided. -- 2020.7.15, Add several related datasets, data annotation and synthesis tools. +- 2020.11.25 Update a new data annotation tool, i.e., [PPOCRLabel](./PPOCRLabel/README_en.md), which is helpful to improve the labeling efficiency. Moreover, the labeling results can be used in training of the PP-OCR system directly. +- 2020.9.22 Update the PP-OCR technical article, https://arxiv.org/abs/2009.09941 +- 2020.9.19 Update the ultra lightweight compressed ppocr_mobile_slim series models, the overall model size is 3.5M (see [PP-OCR Pipeline](#PP-OCR-Pipeline)), suitable for mobile deployment. [Model Downloads](#Supported-Chinese-model-list) +- 2020.9.17 Update the ultra lightweight ppocr_mobile series and general ppocr_server series Chinese and English ocr models, which are comparable to commercial effects. [Model Downloads](#Supported-Chinese-model-list) +- 2020.9.17 update [English recognition model](./doc/doc_en/models_list_en.md#english-recognition-model) and [Multilingual recognition model](doc/doc_en/models_list_en.md#english-recognition-model), `English`, `Chinese`, `German`, `French`, `Japanese` and `Korean` have been supported. Models for more languages will continue to be updated. +- 2020.8.24 Support the use of PaddleOCR through whl package installation，please refer [PaddleOCR Package](./doc/doc_en/whl_en.md) +- 2020.8.21 Update the replay and PPT of the live lesson at Bilibili on August 18, lesson 2, easy to learn and use OCR tool spree. [Get Address](https://aistudio.baidu.com/aistudio/education/group/info/1519) - [more](./doc/doc_en/update_en.md) ## Features -- Ultra-lightweight OCR model, total model size is only 8.6M - - Single model supports Chinese/English numbers combination recognition, vertical text recognition, long text recognition - - Detection model DB (4.1M) + recognition model CRNN (4.5M) -- Various text detection algorithms: EAST, DB -- Various text recognition algorithms: Rosetta, CRNN, STAR-Net, RARE -- Support Linux, Windows, macOS and other systems. +- PPOCR series of high-quality pre-trained models, comparable to commercial effects + - Ultra lightweight ppocr_mobile series models: detection (2.6M) + direction classifier (0.9M) + recognition (4.6M) = 8.1M + - General ppocr_server series models: detection (47.2M) + direction classifier (0.9M) + recognition (107M) = 155.1M + - Ultra lightweight compression ppocr_mobile_slim series models: detection (1.4M) + direction classifier (0.5M) + recognition (1.6M) = 3.5M +- Support Chinese, English, and digit recognition, vertical text recognition, and long text recognition +- Support multi-language recognition: Korean, Japanese, German, French +- Support user-defined training, provides rich predictive inference deployment solutions +- Support PIP installation, easy to use +- Support Linux, Windows, MacOS and other systems ## Visualization -![](doc/imgs_results/11.jpg) +
+ + +
-![](doc/imgs_results/img_10.jpg) +The above pictures are the visualizations of the general ppocr_server model. For more effect pictures, please see [More visualizations](./doc/doc_en/visualization_en.md). -[More visualization](./doc/doc_en/visualization_en.md) + +## Community +- Scan the QR code below with your Wechat, you can access to official technical exchange group. Look forward to your participation. + +
+ +
+ + +## Quick Experience You can also quickly experience the ultra-lightweight OCR : [Online Experience](https://www.paddlepaddle.org.cn/hub/scene/ocr) @@ -42,177 +58,106 @@ Mobile DEMO experience (based on EasyEdge and Paddle-Lite, supports iOS and Andr -### Supported Models: +## PP-OCR 2.0 series model list（Update on Sep 17） + +| Model introduction | Model name | Recommended scene | Detection model | Direction classifier | Recognition model | +| ------------------------------------------------------------ | ---------------------------- | ----------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| Chinese and English ultra-lightweight OCR model (8.1M) | ch_ppocr_mobile_v2.0_xx | Mobile & server |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +| Chinese and English general OCR model (143M) | ch_ppocr_server_v2.0_xx | Server |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_traingit.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | + -|Model Name|Description |Detection Model link|Recognition Model link| Support for space Recognition Model link| -|-|-|-|-|-| -|db_crnn_mobile|ultra-lightweight OCR model|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance_infer.tar) / [pre-train model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance.tar) -|db_crnn_server|General OCR model|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_infer.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance_infer.tar) / [pre-train model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance.tar) +For more model downloads (including multiple languages), please refer to [PP-OCR v2.0 series model downloads](./doc/doc_en/models_list_en.md). +For a new language request, please refer to [Guideline for new language_requests](#language_requests). ## Tutorials - [Installation](./doc/doc_en/installation_en.md) - [Quick Start](./doc/doc_en/quickstart_en.md) -- Algorithm introduction - - [Text Detection Algorithm](#TEXTDETECTIONALGORITHM) - - [Text Recognition Algorithm](#TEXTRECOGNITIONALGORITHM) - - [END-TO-END OCR Algorithm](#ENDENDOCRALGORITHM) -- Model training/evaluation +- [Code Structure](./doc/doc_en/tree_en.md) +- Algorithm Introduction + - [Text Detection Algorithm](./doc/doc_en/algorithm_overview_en.md) + - [Text Recognition Algorithm](./doc/doc_en/algorithm_overview_en.md) + - [PP-OCR Pipeline](#PP-OCR-Pipeline) +- Model Training/Evaluation - [Text Detection](./doc/doc_en/detection_en.md) - [Text Recognition](./doc/doc_en/recognition_en.md) + - [Direction Classification](./doc/doc_en/angle_class_en.md) - [Yml Configuration](./doc/doc_en/config_en.md) - - [Tricks](./doc/doc_en/tricks_en.md) -- Deployment +- Inference and Deployment + - [Quick Inference Based on PIP](./doc/doc_en/whl_en.md) - [Python Inference](./doc/doc_en/inference_en.md) - [C++ Inference](./deploy/cpp_infer/readme_en.md) - - [Serving](./doc/doc_en/serving_en.md) + - [Serving](./deploy/hubserving/readme_en.md) - [Mobile](./deploy/lite/readme_en.md) - - Model Quantization and Compression (coming soon) - - [Benchmark](./doc/doc_en/benchmark_en.md) + - [Model Quantization](./deploy/slim/quantization/README_en.md) + - [Model Compression](./deploy/slim/prune/README_en.md) + - [Benchmark](./doc/doc_en/benchmark_en.md) +- Data Annotation and Synthesis + - [Semi-automatic Annotation Tool](./PPOCRLabel/README_en.md) + - [Data Annotation Tools](./doc/doc_en/data_annotation_en.md) + - [Data Synthesis Tools](./doc/doc_en/data_synthesis_en.md) - Datasets - [General OCR Datasets(Chinese/English)](./doc/doc_en/datasets_en.md) - [HandWritten_OCR_Datasets(Chinese)](./doc/doc_en/handwritten_datasets_en.md) - [Various OCR Datasets(multilingual)](./doc/doc_en/vertical_and_multilingual_datasets_en.md) - - [Data Annotation Tools](./doc/doc_en/data_annotation_en.md) - - [Data Synthesis Tools](./doc/doc_en/data_synthesis_en.md) -- [FAQ](#FAQ) -- Visualization - - [Ultra-lightweight Chinese/English OCR Visualization](#UCOCRVIS) - - [General Chinese/English OCR Visualization](#GeOCRVIS) - - [Chinese/English OCR Visualization (Support Space Recognition )](#SpaceOCRVIS) +- [Visualization](#Visualization) +- [New language requests](#language_requests) +- [FAQ](./doc/doc_en/FAQ_en.md) - [Community](#Community) - [References](./doc/doc_en/reference_en.md) - [License](#LICENSE) - [Contribution](#CONTRIBUTION) - -## Text Detection Algorithm - -PaddleOCR open source text detection algorithms list: -- [x] EAST([paper](https://arxiv.org/abs/1704.03155)) -- [x] DB([paper](https://arxiv.org/abs/1911.08947)) -- [x] SAST([paper](https://arxiv.org/abs/1908.05498))(Baidu Self-Research) - -On the ICDAR2015 dataset, the text detection result is as follows: - -|Model|Backbone|precision|recall|Hmean|Download link| -|-|-|-|-|-|-| -|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[Download link](https://paddleocr.bj.bcebos.com/det_r50_vd_east.tar)| -|EAST|MobileNetV3|81.67%|79.83%|80.74%|[Download link](https://paddleocr.bj.bcebos.com/det_mv3_east.tar)| -|DB|ResNet50_vd|83.79%|80.65%|82.19%|[Download link](https://paddleocr.bj.bcebos.com/det_r50_vd_db.tar)| -|DB|MobileNetV3|75.92%|73.18%|74.53%|[Download link](https://paddleocr.bj.bcebos.com/det_mv3_db.tar)| -|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[Download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_icdar2015.tar)| - -On Total-Text dataset, the text detection result is as follows: - -|Model|Backbone|precision|recall|Hmean|Download link| -|-|-|-|-|-|-| -|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[Download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar)| - -**Note：** Additional data, like icdar2013, icdar2017, COCO-Text, ArT, was added to the model training of SAST. Download English public dataset in organized format used by PaddleOCR from [Baidu Drive](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (download code: 2bpi). - -For use of [LSVT](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_en/datasets_en.md#1-icdar2019-lsvt) street view dataset with a total of 3w training data，the related configuration and pre-trained models for text detection task are as follows: -|Model|Backbone|Configuration file|Pre-trained model| -|-|-|-|-| -|ultra-lightweight OCR model|MobileNetV3|det_mv3_db.yml|[Download link](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db.tar)| -|General OCR model|ResNet50_vd|det_r50_vd_db.yml|[Download link](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db.tar)| + -* Note: For the training and evaluation of the above DB model, post-processing parameters box_thresh=0.6 and unclip_ratio=1.5 need to be set. If using different datasets and different models for training, these two parameters can be adjusted for better result. +## PP-OCR Pipeline -For the training guide and use of PaddleOCR text detection algorithms, please refer to the document [Text detection model training/evaluation/prediction](./doc/doc_en/detection_en.md) - - -## Text Recognition Algorithm - -PaddleOCR open-source text recognition algorithms list: -- [x] CRNN([paper](https://arxiv.org/abs/1507.05717)) -- [x] Rosetta([paper](https://arxiv.org/abs/1910.05085)) -- [x] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) -- [x] RARE([paper](https://arxiv.org/abs/1603.03915v1)) -- [x] SRN([paper](https://arxiv.org/abs/2003.12294))(Baidu Self-Research) - -Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow: - -|Model|Backbone|Avg Accuracy|Module combination|Download link| -|-|-|-|-|-| -|Rosetta|Resnet34_vd|80.24%|rec_r34_vd_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_none_ctc.tar)| -|Rosetta|MobileNetV3|78.16%|rec_mv3_none_none_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_none_none_ctc.tar)| -|CRNN|Resnet34_vd|82.20%|rec_r34_vd_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_bilstm_ctc.tar)| -|CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar)| -|STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar)| -|STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_ctc.tar)| -|RARE|Resnet34_vd|84.90%|rec_r34_vd_tps_bilstm_attn|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_attn.tar)| -|RARE|MobileNetV3|83.32%|rec_mv3_tps_bilstm_attn|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_attn.tar)| -|SRN|Resnet50_vd_fpn|88.33%|rec_r50fpn_vd_none_srn|[Download link](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar)| - -**Note：** SRN model uses data expansion method to expand the two training sets mentioned above, and the expanded data can be downloaded from [Baidu Drive](https://pan.baidu.com/s/1-HSZ-ZVdqBF2HaBZ5pRAKA) (download code: y3ry). - -The average accuracy of the two-stage training in the original paper is 89.74%, and that of one stage training in paddleocr is 88.33%. Both pre-trained weights can be downloaded [here](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar). - -We use [LSVT](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_en/datasets_en.md#1-icdar2019-lsvt) dataset and cropout 30w training data from original photos by using position groundtruth and make some calibration needed. In addition, based on the LSVT corpus, 500w synthetic data is generated to train the model. The related configuration and pre-trained models are as follows: - -|Model|Backbone|Configuration file|Pre-trained model| -|-|-|-|-| -|ultra-lightweight OCR model|MobileNetV3|rec_chinese_lite_train.yml|[Download link](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance_infer.tar) & [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance.tar)| -|General OCR model|Resnet34_vd|rec_chinese_common_train.yml|[Download link](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn.tar)|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance_infer.tar) & [pre-trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance.tar)| - -Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./doc/doc_en/recognition_en.md) +
+ +
- -## END-TO-END OCR Algorithm -- [ ] [End2End-PSL](https://arxiv.org/abs/1909.07808)(Baidu Self-Research, coming soon) +PP-OCR is a practical ultra-lightweight OCR system. It is mainly composed of three parts: DB text detection, detection frame correction and CRNN text recognition. The system adopts 19 effective strategies from 8 aspects including backbone network selection and adjustment, prediction head design, data augmentation, learning rate transformation strategy, regularization parameter selection, pre-training model use, and automatic model tailoring and quantization to optimize and slim down the models of each module. The final results are an ultra-lightweight Chinese and English OCR model with an overall size of 3.5M and a 2.8M English digital OCR model. For more details, please refer to the PP-OCR technical article (https://arxiv.org/abs/2009.09941). Besides, The implementation of the FPGM Pruner and PACT quantization is based on [PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim). -## Visualization - -### 1.Ultra-lightweight Chinese/English OCR Visualization [more](./doc/doc_en/visualization_en.md) +## Visualization [more](./doc/doc_en/visualization_en.md) +- Chinese OCR model
- + + + +
- -### 2. General Chinese/English OCR Visualization [more](./doc/doc_en/visualization_en.md) - +- English OCR model
- +
- -### 3.Chinese/English OCR Visualization (Space_support) [more](./doc/doc_en/visualization_en.md) - +- Multilingual OCR model
- + +
- - -## FAQ -1. Error when using attention-based recognition model: KeyError: 'predict' - - The inference of recognition model based on attention loss is still being debugged. For Chinese text recognition, it is recommended to choose the recognition model based on CTC loss first. In practice, it is also found that the recognition model based on attention loss is not as effective as the one based on CTC loss. - -2. About inference speed - When there are a lot of texts in the picture, the prediction time will increase. You can use `--rec_batch_num` to set a smaller prediction batch size. The default value is 30, which can be changed to 10 or other values. + +## Guideline for new language requests -3. Service deployment and mobile deployment +If you want to request a new language support, a PR with 2 following files are needed： - It is expected that the service deployment based on Serving and the mobile deployment based on Paddle Lite will be released successively in mid-to-late June. Stay tuned for more updates. +1. In folder [ppocr/utils/dict](https://github.com/PaddlePaddle/PaddleOCR/tree/develop/ppocr/utils/dict), +it is necessary to submit the dict text to this path and name it with `{language}_dict.txt` that contains a list of all characters. Please see the format example from other files in that folder. -4. Release time of self-developed algorithm +2. In folder [ppocr/utils/corpus](https://github.com/PaddlePaddle/PaddleOCR/tree/develop/ppocr/utils/corpus), +it is necessary to submit the corpus to this path and name it with `{language}_corpus.txt` that contains a list of words in your language. +Maybe, 50000 words per language is necessary at least. +Of course, the more, the better. - Baidu Self-developed algorithms such as SAST, SRN and end2end PSL will be released in June or July. Please be patient. +If your language has unique elements, please tell me in advance within any way, such as useful links, wikipedia and so on. -[more](./doc/doc_en/FAQ_en.md) +More details, please refer to [Multilingual OCR Development Plan](https://github.com/PaddlePaddle/PaddleOCR/issues/1048). - -## Community -Scan the QR code below with your wechat and completing the questionnaire, you can access to offical technical exchange group. - -
- -
## License @@ -229,3 +174,7 @@ We welcome all the contributions to PaddleOCR and appreciate for your feedback v - Thanks [authorfu](https://github.com/authorfu) for contributing Android demo and [xiadeye](https://github.com/xiadeye) contributing iOS demo, respectively. - Thanks [BeyondYourself](https://github.com/BeyondYourself) for contributing many great suggestions and simplifying part of the code style. - Thanks [tangmq](https://gitee.com/tangmq) for contributing Dockerized deployment services to PaddleOCR and supporting the rapid release of callable Restful API services. +- Thanks [lijinhan](https://github.com/lijinhan) for contributing a new way, i.e., java SpringBoot, to achieve the request for the Hubserving deployment. +- Thanks [Mejans](https://github.com/Mejans) for contributing the Occitan corpus and character set. +- Thanks [LKKlein](https://github.com/LKKlein) for contributing a new deploying package with the Golang program language. +- Thanks [Evezerest](https://github.com/Evezerest), [ninetailskim](https://github.com/ninetailskim), [edencfc](https://github.com/edencfc), [BeyondYourself](https://github.com/BeyondYourself) and [1084667371](https://github.com/1084667371) for contributing a new data annotation tool, i.e., PPOCRLabel。 diff --git a/configs/cls/cls_mv3.yml b/configs/cls/cls_mv3.yml index c2b171590cfe2d6308bbee16e534bfc21099309e..b165bc4830f01f0e63c43b1d6a9635e432dc7605 100644 --- a/configs/cls/cls_mv3.yml +++ b/configs/cls/cls_mv3.yml @@ -8,7 +8,6 @@ Global: # evaluation is run every 5000 iterations after the 4000th iteration eval_batch_step: [0, 1000] # if pretrained_model is saved in static mode, load_static_weights must set to True - load_static_weights: True cal_metric_during_train: True pretrained_model: checkpoints: diff --git a/configs/det/bak/det_r50_vd_db.yml b/configs/det/bak/det_r50_vd_db.yml deleted file mode 100644 index a07273b4ae294164c0c5d8166ec602beade55259..0000000000000000000000000000000000000000 --- a/configs/det/bak/det_r50_vd_db.yml +++ /dev/null @@ -1,130 +0,0 @@ -Global: - use_gpu: true - epoch_num: 1200 - log_smooth_window: 20 - print_batch_step: 2 - save_model_dir: ./output/det_r50_vd/ - save_epoch_step: 1200 - # evaluation is run every 5000 iterations after the 4000th iteration - eval_batch_step: 8 - # if pretrained_model is saved in static mode, load_static_weights must set to True - load_static_weights: True - cal_metric_during_train: False - pretrained_model: ./pretrain_models/ResNet50_vd_ssld_pretrained/ - checkpoints: - save_inference_dir: - use_visualdl: True - infer_img: doc/imgs_en/img_10.jpg - save_res_path: ./output/det_db/predicts_db.txt - -Optimizer: - name: Adam - beta1: 0.9 - beta2: 0.999 - learning_rate: - lr: 0.001 - regularizer: - name: 'L2' - factor: 0 - -Architecture: - type: det - algorithm: DB - Transform: - Backbone: - name: ResNet - layers: 50 - Neck: - name: FPN - out_channels: 256 - Head: - name: DBHead - k: 50 - -Loss: - name: DBLoss - balance_loss: true - main_loss_type: DiceLoss - alpha: 5 - beta: 10 - ohem_ratio: 3 - -PostProcess: - name: DBPostProcess - thresh: 0.3 - box_thresh: 0.6 - max_candidates: 1000 - unclip_ratio: 1.5 - -Metric: - name: DetMetric - main_indicator: hmean - -TRAIN: - dataset: - name: SimpleDataSet - data_dir: ./detection/ - file_list: - - ./detection/train_icdar2015_label.txt # dataset1 - ratio_list: [1.0] - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - DetLabelEncode: # Class handling label - - IaaAugment: - augmenter_args: - - { 'type': Fliplr, 'args': { 'p': 0.5 } } - - { 'type': Affine, 'args': { 'rotate': [ -10,10 ] } } - - { 'type': Resize,'args': { 'size': [ 0.5,3 ] } } - - EastRandomCropData: - size: [ 640,640 ] - max_tries: 50 - keep_ratio: true - - MakeBorderMap: - shrink_ratio: 0.4 - thresh_min: 0.3 - thresh_max: 0.7 - - MakeShrinkMap: - shrink_ratio: 0.4 - min_text_size: 8 - - NormalizeImage: - scale: 1./255. - mean: [ 0.485, 0.456, 0.406 ] - std: [ 0.229, 0.224, 0.225 ] - order: 'hwc' - - ToCHWImage: - - keepKeys: - keep_keys: ['image','threshold_map','threshold_mask','shrink_map','shrink_mask'] # dataloader will return list in this order - loader: - shuffle: True - drop_last: False - batch_size: 16 - num_workers: 8 - -EVAL: - dataset: - name: SimpleDataSet - data_dir: ./detection/ - file_list: - - ./detection/test_icdar2015_label.txt - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - DetLabelEncode: # Class handling label - - DetResizeForTest: - image_shape: [736,1280] - - NormalizeImage: - scale: 1./255. - mean: [ 0.485, 0.456, 0.406 ] - std: [ 0.229, 0.224, 0.225 ] - order: 'hwc' - - ToCHWImage: - - keepKeys: - keep_keys: ['image','shape','polys','ignore_tags'] - loader: - shuffle: False - drop_last: False - batch_size: 1 # must be 1 - num_workers: 8 \ No newline at end of file diff --git a/configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml b/configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml index 275c71b97d21e6b168ef6aafae67eb2eb91c6f2b..fd88495928b18c300386c6a9fd0cf57d840db21e 100644 --- a/configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml +++ b/configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml @@ -11,7 +11,7 @@ Global: load_static_weights: True cal_metric_during_train: False pretrained_model: ./pretrain_models/MobileNetV3_large_x0_5_pretrained - checkpoints: #./output/det_db_0.001_DiceLoss_256_pp_config_2.0b_4gpu/best_accuracy + checkpoints: save_inference_dir: use_visualdl: False infer_img: doc/imgs_en/img_10.jpg diff --git a/configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml b/configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml index e34d94490fbfa81452a7ff507f6568e5ac29e3b7..2694601254935be7d003148681334263d734579a 100644 --- a/configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml +++ b/configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml @@ -11,7 +11,7 @@ Global: load_static_weights: True cal_metric_during_train: False pretrained_model: ./pretrain_models/ResNet18_vd_pretrained - checkpoints: #./output/det_db_0.001_DiceLoss_256_pp_config_2.0b_4gpu/best_accuracy + checkpoints: save_inference_dir: use_visualdl: False infer_img: doc/imgs_en/img_10.jpg diff --git a/configs/det/det_mv3_db.yml b/configs/det/det_mv3_db.yml index 640f3a205b1fd3ec7fe19d5c6b6e3aef9ddf3968..36a6f755383e525a8a496b060465cf027f3f31f8 100644 --- a/configs/det/det_mv3_db.yml +++ b/configs/det/det_mv3_db.yml @@ -11,7 +11,7 @@ Global: load_static_weights: True cal_metric_during_train: False pretrained_model: ./pretrain_models/MobileNetV3_large_x0_5_pretrained - checkpoints: #./output/det_db_0.001_DiceLoss_256_pp_config_2.0b_4gpu/best_accuracy + checkpoints: save_inference_dir: use_visualdl: False infer_img: doc/imgs_en/img_10.jpg diff --git a/configs/det/det_r50_vd_db.yml b/configs/det/det_r50_vd_db.yml index 491983f57a59f0a4105712743a69a56c8212a0e9..b70ab7505a4c1970f084b6c02233526d7f7188b9 100644 --- a/configs/det/det_r50_vd_db.yml +++ b/configs/det/det_r50_vd_db.yml @@ -3,7 +3,7 @@ Global: epoch_num: 1200 log_smooth_window: 20 print_batch_step: 10 - save_model_dir: ./output/det_rc/det_r50_vd/ + save_model_dir: ./output/det_r50_vd/ save_epoch_step: 1200 # evaluation is run every 5000 iterations after the 4000th iteration eval_batch_step: [5000,4000] diff --git a/configs/rec/bak/rec_mv3_none_bilstm_ctc_simple.yml b/configs/rec/bak/rec_mv3_none_bilstm_ctc_simple.yml deleted file mode 100644 index 1be7512c9d793b38b7d5c23ab4e55972e793c28b..0000000000000000000000000000000000000000 --- a/configs/rec/bak/rec_mv3_none_bilstm_ctc_simple.yml +++ /dev/null @@ -1,106 +0,0 @@ -Global: - use_gpu: false - epoch_num: 500 - log_smooth_window: 20 - print_batch_step: 10 - save_model_dir: ./output/rec/mv3_none_bilstm_ctc/ - save_epoch_step: 500 - # evaluation is run every 5000 iterations after the 4000th iteration - eval_batch_step: 127 - # if pretrained_model is saved in static mode, load_static_weights must set to True - load_static_weights: True - cal_metric_during_train: True - pretrained_model: - checkpoints: - save_inference_dir: - use_visualdl: False - infer_img: doc/imgs_words/ch/word_1.jpg - # for data or label process - max_text_length: 80 - character_dict_path: ppocr/utils/ppocr_keys_v1.txt - character_type: 'ch' - use_space_char: False - infer_mode: False - use_tps: False - - -Optimizer: - name: Adam - beta1: 0.9 - beta2: 0.999 - learning_rate: - lr: 0.001 - regularizer: - name: 'L2' - factor: 0.00001 - -Architecture: - type: rec - algorithm: CRNN - Transform: - Backbone: - name: MobileNetV3 - scale: 0.5 - model_name: small - small_stride: [ 1, 2, 2, 2 ] - Neck: - name: SequenceEncoder - encoder_type: fc - hidden_size: 96 - Head: - name: CTC - fc_decay: 0.00001 - -Loss: - name: CTCLoss - -PostProcess: - name: CTCLabelDecode - -Metric: - name: RecMetric - main_indicator: acc - -TRAIN: - dataset: - name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/train.txt # dataset1 - ratio_list: [ 0.4,0.6 ] - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - CTCLabelEncode: # Class handling label - - RecAug: - - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order - loader: - batch_size: 256 - shuffle: True - drop_last: True - num_workers: 8 - -EVAL: - dataset: - name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/val.txt - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - CTCLabelEncode: # Class handling label - - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order - loader: - shuffle: False - drop_last: False - batch_size: 256 - num_workers: 8 diff --git a/configs/rec/bak/rec_r34_vd_none_bilstm_ctc.yml b/configs/rec/bak/rec_r34_vd_none_bilstm_ctc.yml deleted file mode 100644 index 36e3d1c81cb5e5ad744576dc6d454e8f31d965dc..0000000000000000000000000000000000000000 --- a/configs/rec/bak/rec_r34_vd_none_bilstm_ctc.yml +++ /dev/null @@ -1,104 +0,0 @@ -Global: - use_gpu: false - epoch_num: 500 - log_smooth_window: 20 - print_batch_step: 10 - save_model_dir: ./output/rec/res34_none_bilstm_ctc/ - save_epoch_step: 500 - # evaluation is run every 5000 iterations after the 4000th iteration - eval_batch_step: 127 - # if pretrained_model is saved in static mode, load_static_weights must set to True - load_static_weights: True - cal_metric_during_train: True - pretrained_model: - checkpoints: - save_inference_dir: - use_visualdl: False - infer_img: doc/imgs_words/ch/word_1.jpg - # for data or label process - max_text_length: 80 - character_dict_path: ppocr/utils/ppocr_keys_v1.txt - character_type: 'ch' - use_space_char: False - infer_mode: False - use_tps: False - - -Optimizer: - name: Adam - beta1: 0.9 - beta2: 0.999 - learning_rate: - lr: 0.001 - regularizer: - name: 'L2' - factor: 0.00001 - -Architecture: - type: rec - algorithm: CRNN - Transform: - Backbone: - name: ResNet - layers: 34 - Neck: - name: SequenceEncoder - encoder_type: fc - hidden_size: 96 - Head: - name: CTC - fc_decay: 0.00001 - -Loss: - name: CTCLoss - -PostProcess: - name: CTCLabelDecode - -Metric: - name: RecMetric - main_indicator: acc - -TRAIN: - dataset: - name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/train.txt # dataset1 - ratio_list: [ 0.4,0.6 ] - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - CTCLabelEncode: # Class handling label - - RecAug: - - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order - loader: - batch_size: 256 - shuffle: True - drop_last: True - num_workers: 8 - -EVAL: - dataset: - name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/val.txt - transforms: - - DecodeImage: # load image - img_mode: BGR - channel_first: False - - CTCLabelEncode: # Class handling label - - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order - loader: - shuffle: False - drop_last: False - batch_size: 256 - num_workers: 8 diff --git a/configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yaml b/configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml similarity index 97% rename from configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yaml rename to configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml index 6d53ce8b04003612f10ed6abf3866474261cc812..1db3e1cb8633f03f91d1d44064a19f7661e57b12 100644 --- a/configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yaml +++ b/configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml @@ -3,7 +3,7 @@ Global: epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 - save_model_dir: ./output/rec_chinese_common_v1.1 + save_model_dir: ./output/rec_chinese_common_v2.0 save_epoch_step: 3 # evaluation is run every 5000 iterations after the 4000th iteration eval_batch_step: [0, 2000] diff --git a/configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yaml b/configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml similarity index 96% rename from configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yaml rename to configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml index 94a22e5c6e738429939d8df7b9d1c556abba7f6c..dc9d650f30f3d1086616a81c27aaf5db389a1fe7 100644 --- a/configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yaml +++ b/configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml @@ -3,7 +3,7 @@ Global: epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 - save_model_dir: ./output/rec_chinese_lite_v1.1 + save_model_dir: ./output/rec_chinese_lite_v2.0 save_epoch_step: 3 # evaluation is run every 5000 iterations after the 4000th iteration eval_batch_step: [0, 2000] @@ -19,7 +19,7 @@ Global: character_type: ch max_text_length: 25 infer_mode: False - use_space_char: False + use_space_char: True Optimizer: diff --git a/configs/rec/multi_language/rec_en_number_lite_train.yml b/configs/rec/multi_language/rec_en_number_lite_train.yml index 70d825e61ab48029382426b0ec56b4a1a4ac439d..cee0512114fe9d488004a71cf6f0a0409822a4b5 100644 --- a/configs/rec/multi_language/rec_en_number_lite_train.yml +++ b/configs/rec/multi_language/rec_en_number_lite_train.yml @@ -1,5 +1,5 @@ Global: - use_gpu: true + use_gpu: True epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 @@ -15,7 +15,7 @@ Global: use_visualdl: False infer_img: # for data or label process - character_dict_path: ppocr/utils/dict/ic15_dict.txt + character_dict_path: ppocr/utils/dict/en_dict.txt character_type: ch max_text_length: 25 infer_mode: False diff --git a/configs/rec/multi_language/rec_french_lite_train.yml b/configs/rec/multi_language/rec_french_lite_train.yml index 0e8f4eb3a803e8958cf38f0b55fca25e7204bd9c..63378d38a0d31fc77c33173e0ed864f28c5c3a8b 100644 --- a/configs/rec/multi_language/rec_french_lite_train.yml +++ b/configs/rec/multi_language/rec_french_lite_train.yml @@ -1,5 +1,5 @@ Global: - use_gpu: true + use_gpu: True epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 @@ -9,9 +9,9 @@ Global: eval_batch_step: [0, 2000] # if pretrained_model is saved in static mode, load_static_weights must set to True cal_metric_during_train: True - pretrained_model: + pretrained_model: checkpoints: - save_inference_dir: + save_inference_dir: use_visualdl: False infer_img: # for data or label process @@ -19,7 +19,7 @@ Global: character_type: french max_text_length: 25 infer_mode: False - use_space_char: True + use_space_char: False Optimizer: diff --git a/configs/rec/multi_language/rec_german_lite_train.yml b/configs/rec/multi_language/rec_german_lite_train.yml index 9978a21e1e8c70e44e6fd670eca305d045a919e9..1651510c5e4597e82298135d2f6c64aa747cf961 100644 --- a/configs/rec/multi_language/rec_german_lite_train.yml +++ b/configs/rec/multi_language/rec_german_lite_train.yml @@ -1,5 +1,5 @@ Global: - use_gpu: true + use_gpu: True epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 @@ -19,7 +19,7 @@ Global: character_type: german max_text_length: 25 infer_mode: False - use_space_char: True + use_space_char: False Optimizer: diff --git a/configs/rec/multi_language/rec_japan_lite_train.yml b/configs/rec/multi_language/rec_japan_lite_train.yml index 938d377e5d6497bea72fa27b0adcd3d0d691d431..bb47584edbc70f68d8d2d89dced3ec9b12f0e1cb 100644 --- a/configs/rec/multi_language/rec_japan_lite_train.yml +++ b/configs/rec/multi_language/rec_japan_lite_train.yml @@ -1,5 +1,5 @@ Global: - use_gpu: true + use_gpu: True epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 @@ -19,7 +19,7 @@ Global: character_type: japan max_text_length: 25 infer_mode: False - use_space_char: True + use_space_char: False Optimizer: diff --git a/configs/rec/multi_language/rec_korean_lite_train.yml b/configs/rec/multi_language/rec_korean_lite_train.yml index 7b070c449e9511e353b05877eaf217d0b286b47f..77f15524f78cd7f1c3dcf4988960e718422f5d89 100644 --- a/configs/rec/multi_language/rec_korean_lite_train.yml +++ b/configs/rec/multi_language/rec_korean_lite_train.yml @@ -1,5 +1,5 @@ Global: - use_gpu: true + use_gpu: True epoch_num: 500 log_smooth_window: 20 print_batch_step: 10 @@ -19,7 +19,7 @@ Global: character_type: korean max_text_length: 25 infer_mode: False - use_space_char: True + use_space_char: False Optimizer: diff --git a/configs/rec/bak/rec_r34_vd_none_none_ctc.yml b/configs/rec/rec_icdar15_train.yml similarity index 55% rename from configs/rec/bak/rec_r34_vd_none_none_ctc.yml rename to configs/rec/rec_icdar15_train.yml index 641e855b431e459536453275759c6a5f064c15fb..7efbd5cf0d963229a94aa43558589b828d17cbd0 100644 --- a/configs/rec/bak/rec_r34_vd_none_none_ctc.yml +++ b/configs/rec/rec_icdar15_train.yml @@ -1,41 +1,38 @@ Global: - use_gpu: false - epoch_num: 500 + use_gpu: true + epoch_num: 72 log_smooth_window: 20 print_batch_step: 10 - save_model_dir: ./output/rec/res34_none_none_ctc/ - save_epoch_step: 500 - # evaluation is run every 5000 iterations after the 4000th iteration - eval_batch_step: 127 + save_model_dir: ./output/rec/ic15/ + save_epoch_step: 3 + # evaluation is run every 2000 iterations + eval_batch_step: [0, 2000] # if pretrained_model is saved in static mode, load_static_weights must set to True - load_static_weights: True cal_metric_during_train: True pretrained_model: checkpoints: save_inference_dir: use_visualdl: False - infer_img: doc/imgs_words/ch/word_1.jpg + infer_img: doc/imgs_words_en/word_10.png # for data or label process - max_text_length: 80 - character_dict_path: ppocr/utils/ppocr_keys_v1.txt - character_type: 'ch' - use_space_char: False + character_dict_path: ppocr/utils/ic15_dict.txt + character_type: ch + max_text_length: 25 infer_mode: False - use_tps: False - + use_space_char: False Optimizer: name: Adam beta1: 0.9 beta2: 0.999 - learning_rate: - lr: 0.001 + lr: + learning_rate: 0.0005 regularizer: name: 'L2' - factor: 0.00001 + factor: 0 Architecture: - type: rec + model_type: rec algorithm: CRNN Transform: Backbone: @@ -43,10 +40,11 @@ Architecture: layers: 34 Neck: name: SequenceEncoder - encoder_type: reshape + encoder_type: rnn + hidden_size: 256 Head: - name: CTC - fc_decay: 0.00001 + name: CTCHead + fc_decay: 0 Loss: name: CTCLoss @@ -58,46 +56,42 @@ Metric: name: RecMetric main_indicator: acc -TRAIN: +Train: dataset: name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/train.txt # dataset1 - ratio_list: [ 0.4,0.6 ] + data_dir: ./train_data/ + label_file_list: ["./train_data/train_list.txt"] transforms: - DecodeImage: # load image img_mode: BGR channel_first: False - CTCLabelEncode: # Class handling label - - RecAug: - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order + image_shape: [3, 32, 100] + - KeepKeys: + keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order loader: - batch_size: 256 shuffle: True + batch_size_per_card: 256 drop_last: True num_workers: 8 -EVAL: +Eval: dataset: name: SimpleDataSet - data_dir: ./rec - file_list: - - ./rec/val.txt + data_dir: ./train_data/ + label_file_list: ["./train_data/train_list.txt"] transforms: - DecodeImage: # load image img_mode: BGR channel_first: False - CTCLabelEncode: # Class handling label - RecResizeImg: - image_shape: [ 3,32,320 ] - - keepKeys: - keep_keys: [ 'image','label','length' ] # dataloader will return list in this order + image_shape: [3, 32, 100] + - KeepKeys: + keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order loader: shuffle: False drop_last: False - batch_size: 256 - num_workers: 8 + batch_size_per_card: 256 + num_workers: 4 diff --git a/deploy/cpp_infer/src/ocr_cls.cpp b/deploy/cpp_infer/src/ocr_cls.cpp index 40debaa7835d3174627f8b0528abba673c6e3d86..fed2023f9f111294a07a9c841f4843404bbd9af2 100644 --- a/deploy/cpp_infer/src/ocr_cls.cpp +++ b/deploy/cpp_infer/src/ocr_cls.cpp @@ -81,7 +81,8 @@ cv::Mat Classifier::Run(cv::Mat &img) { void Classifier::LoadModel(const std::string &model_dir) { AnalysisConfig config; - config.SetModel(model_dir + "/model", model_dir + "/params"); + config.SetModel(model_dir + "/inference.pdmodel", + model_dir + "/inference.pdiparams"); if (this->use_gpu_) { config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); diff --git a/deploy/cpp_infer/src/ocr_det.cpp b/deploy/cpp_infer/src/ocr_det.cpp index 1e1aaa1bf04d5f0bb80163d39d3a74f0312ae5af..e253f9cc89810f4d1adfca5be5186220a873d1a2 100644 --- a/deploy/cpp_infer/src/ocr_det.cpp +++ b/deploy/cpp_infer/src/ocr_det.cpp @@ -18,7 +18,8 @@ namespace PaddleOCR { void DBDetector::LoadModel(const std::string &model_dir) { AnalysisConfig config; - config.SetModel(model_dir + "/model", model_dir + "/params"); + config.SetModel(model_dir + "/inference.pdmodel", + model_dir + "/inference.pdiparams"); if (this->use_gpu_) { config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); diff --git a/deploy/cpp_infer/src/ocr_rec.cpp b/deploy/cpp_infer/src/ocr_rec.cpp index 009b6b75f53230c781dee285774da369ae10ce4b..d4deb5a17fc47427eb92cda02c270d268cfcafc7 100644 --- a/deploy/cpp_infer/src/ocr_rec.cpp +++ b/deploy/cpp_infer/src/ocr_rec.cpp @@ -103,7 +103,8 @@ void CRNNRecognizer::Run(std::vector>> boxes, void CRNNRecognizer::LoadModel(const std::string &model_dir) { AnalysisConfig config; - config.SetModel(model_dir + "/model", model_dir + "/params"); + config.SetModel(model_dir + "/inference.pdmodel", + model_dir + "/inference.pdiparams"); if (this->use_gpu_) { config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); diff --git a/deploy/docker/hubserving/README.md b/deploy/docker/hubserving/README.md new file mode 100644 index 0000000000000000000000000000000000000000..d4db277ffbeaf1efba18c0caef550404e08f2e85 --- /dev/null +++ b/deploy/docker/hubserving/README.md @@ -0,0 +1,54 @@ +English | [简体中文](README_cn.md) + +## Introduction +Many users hope package the PaddleOCR service into a docker image, so that it can be quickly released and used in the docker or k8s environment. + +This page provides some standardized code to achieve this goal. You can quickly publish the PaddleOCR project into a callable Restful API service through the following steps. (At present, the deployment based on the HubServing mode is implemented first, and author plans to increase the deployment of the PaddleServing mode in the futrue) + +## 1. Prerequisites + +You need to install the following basic components first： +a. Docker +b. Graphics driver and CUDA 10.0+（GPU） +c. NVIDIA Container Toolkit（GPU，Docker 19.03+ can skip this） +d. cuDNN 7.6+（GPU） + +## 2. Build Image +a. Goto Dockerfile directory（ps：Need to distinguish between cpu and gpu version, the following takes cpu as an example, gpu version needs to replace the keyword） +``` +cd deploy/docker/hubserving/cpu +``` +c. Build image +``` +docker build -t paddleocr:cpu . +``` + +## 3. Start container +a. CPU version +``` +sudo docker run -dp 8868:8868 --name paddle_ocr paddleocr:cpu +``` +b. GPU version (base on NVIDIA Container Toolkit) +``` +sudo nvidia-docker run -dp 8868:8868 --name paddle_ocr paddleocr:gpu +``` +c. GPU version (Docker 19.03++) +``` +sudo docker run -dp 8868:8868 --gpus all --name paddle_ocr paddleocr:gpu +``` +d. Check service status（If you can see the following statement then it means completed：Successfully installed ocr_system && Running on http://0.0.0.0:8868/） +``` +docker logs -f paddle_ocr +``` + +## 4. Test +a. Calculate the Base64 encoding of the picture to be recognized (if you just test, you can use a free online tool, like：https://freeonlinetools24.com/base64-image/） +b. Post a service request（sample request in sample_request.txt） + +``` +curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"Input image Base64 encode(need to delete the code 'data:image/jpg;base64,'）\"]}" http://localhost:8868/predict/ocr_system +``` +c. Get resposne（If the call is successful, the following result will be returned） +``` +{"msg":"","results":[[{"confidence":0.8403433561325073,"text":"约定","text_region":[[345,377],[641,390],[634,540],[339,528]]},{"confidence":0.8131805658340454,"text":"最终相遇","text_region":[[356,532],[624,530],[624,596],[356,598]]}]],"status":"0"} +``` diff --git a/deploy/docker/hubserving/README_cn.md b/deploy/docker/hubserving/README_cn.md new file mode 100644 index 0000000000000000000000000000000000000000..046903c4c7d80327ef9e8537e735d3b87104dddf --- /dev/null +++ b/deploy/docker/hubserving/README_cn.md @@ -0,0 +1,53 @@ +[English](README.md) | 简体中文 + +## Docker化部署服务 +在日常项目应用中，相信大家一般都会希望能通过Docker技术，把PaddleOCR服务打包成一个镜像，以便在Docker或k8s环境里，快速发布上线使用。 + +本文将提供一些标准化的代码来实现这样的目标。大家通过如下步骤可以把PaddleOCR项目快速发布成可调用的Restful API服务。（目前暂时先实现了基于HubServing模式的部署，后续作者计划增加PaddleServing模式的部署） + +## 1.实施前提准备 + +需要先完成如下基本组件的安装： +a. Docker环境 +b. 显卡驱动和CUDA 10.0+（GPU） +c. NVIDIA Container Toolkit（GPU，Docker 19.03以上版本可以跳过此步） +d. cuDNN 7.6+（GPU） + +## 2.制作镜像 +a.切换至Dockerfile目录（注：需要区分cpu或gpu版本，下文以cpu为例，gpu版本需要替换一下关键字即可） +``` +cd deploy/docker/hubserving/cpu +``` +c.生成镜像 +``` +docker build -t paddleocr:cpu . +``` + +## 3.启动Docker容器 +a. CPU 版本 +``` +sudo docker run -dp 8868:8868 --name paddle_ocr paddleocr:cpu +``` +b. GPU 版本 (通过NVIDIA Container Toolkit) +``` +sudo nvidia-docker run -dp 8868:8868 --name paddle_ocr paddleocr:gpu +``` +c. GPU 版本 (Docker 19.03以上版本，可以直接用如下命令) +``` +sudo docker run -dp 8868:8869 --gpus all --name paddle_ocr paddleocr:gpu +``` +d. 检查服务运行情况（出现：Successfully installed ocr_system和Running on http://0.0.0.0:8868 等信息，表示运行成功） +``` +docker logs -f paddle_ocr +``` + +## 4.测试服务 +a. 计算待识别图片的Base64编码（如果只是测试一下效果，可以通过免费的在线工具实现，如：http://tool.chinaz.com/tools/imgtobase/） +b. 发送服务请求（可参见sample_request.txt中的值） +``` +curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"填入图片Base64编码(需要删除'data:image/jpg;base64,'）\"]}" http://localhost:8868/predict/ocr_system +``` +c. 返回结果（如果调用成功，会返回如下结果） +``` +{"msg":"","results":[[{"confidence":0.8403433561325073,"text":"约定","text_region":[[345,377],[641,390],[634,540],[339,528]]},{"confidence":0.8131805658340454,"text":"最终相遇","text_region":[[356,532],[624,530],[624,596],[356,598]]}]],"status":"0"} +``` diff --git a/deploy/docker/hubserving/cpu/Dockerfile b/deploy/docker/hubserving/cpu/Dockerfile new file mode 100644 index 0000000000000000000000000000000000000000..342f1e819a24721719566b9f3cfc81666a225b9b --- /dev/null +++ b/deploy/docker/hubserving/cpu/Dockerfile @@ -0,0 +1,32 @@ +# Version: 1.0.0 +FROM hub.baidubce.com/paddlepaddle/paddle:latest-gpu-cuda10.0-cudnn7-dev + +# PaddleOCR base on Python3.7 +RUN pip3.7 install --upgrade pip -i https://mirror.baidu.com/pypi/simple + +RUN python3.7 -m pip install paddlepaddle==2.0.0rc0 -i https://mirror.baidu.com/pypi/simple + +RUN pip3.7 install paddlehub --upgrade -i https://mirror.baidu.com/pypi/simple + +RUN git clone https://github.com/PaddlePaddle/PaddleOCR.git /PaddleOCR + +WORKDIR /PaddleOCR + +RUN pip3.7 install -r requirements.txt -i https://mirror.baidu.com/pypi/simple + +RUN mkdir -p /PaddleOCR/inference/ +# Download orc detect model(light version). if you want to change normal version, you can change ch_ppocr_mobile_v1.1_det_infer to ch_ppocr_server_v1.1_det_infer, also remember change det_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file} -C /PaddleOCR/inference/ + +# Download direction classifier(light version). If you want to change normal version, you can change ch_ppocr_mobile_v1.1_cls_infer to ch_ppocr_mobile_v1.1_cls_infer, also remember change cls_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file}.tar -C /PaddleOCR/inference/ + +# Download orc recognition model(light version). If you want to change normal version, you can change ch_ppocr_mobile_v1.1_rec_infer to ch_ppocr_server_v1.1_rec_infer, also remember change rec_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file}.tar -C /PaddleOCR/inference/ + +EXPOSE 8868 + +CMD ["/bin/bash","-c","hub install deploy/hubserving/ocr_system/ && hub serving start -m ocr_system"] \ No newline at end of file diff --git a/deploy/docker/hubserving/gpu/Dockerfile b/deploy/docker/hubserving/gpu/Dockerfile new file mode 100644 index 0000000000000000000000000000000000000000..222d053d953f64d2b7f2d2c0b975ba7169426d92 --- /dev/null +++ b/deploy/docker/hubserving/gpu/Dockerfile @@ -0,0 +1,32 @@ +# Version: 1.0.0 +FROM hub.baidubce.com/paddlepaddle/paddle:latest-gpu-cuda10.0-cudnn7-dev + +# PaddleOCR base on Python3.7 +RUN pip3.7 install --upgrade pip -i https://mirror.baidu.com/pypi/simple + +RUN python3.7 -m pip install paddlepaddle-gpu==2.0.0rc0 -i https://mirror.baidu.com/pypi/simple + +RUN pip3.7 install paddlehub --upgrade -i https://mirror.baidu.com/pypi/simple + +RUN git clone https://github.com/PaddlePaddle/PaddleOCR.git /PaddleOCR + +WORKDIR /PaddleOCR + +RUN pip3.7 install -r requirements.txt -i https://mirror.baidu.com/pypi/simple + +RUN mkdir -p /PaddleOCR/inference/ +# Download orc detect model(light version). if you want to change normal version, you can change ch_ppocr_mobile_v1.1_det_infer to ch_ppocr_server_v1.1_det_infer, also remember change det_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file}.tar -C /PaddleOCR/inference/ + +# Download direction classifier(light version). If you want to change normal version, you can change ch_ppocr_mobile_v1.1_cls_infer to ch_ppocr_mobile_v1.1_cls_infer, also remember change cls_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file} -C /PaddleOCR/inference/ + +# Download orc recognition model(light version). If you want to change normal version, you can change ch_ppocr_mobile_v1.1_rec_infer to ch_ppocr_server_v1.1_rec_infer, also remember change rec_model_dir in deploy/hubserving/ocr_system/params.py） +ADD {link} /PaddleOCR/inference/ +RUN tar xf /PaddleOCR/inference/{file}.tar -C /PaddleOCR/inference/ + +EXPOSE 8868 + +CMD ["/bin/bash","-c","hub install deploy/hubserving/ocr_system/ && hub serving start -m ocr_system"] \ No newline at end of file diff --git a/deploy/docker/hubserving/sample_request.txt b/deploy/docker/hubserving/sample_request.txt new file mode 100644 index 0000000000000000000000000000000000000000..ec2b25b1f54cb7d7cdaeaff6162810a5772442c6 --- /dev/null +++ b/deploy/docker/hubserving/sample_request.txt @@ -0,0 +1 @@ +curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"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\"]}" http://localhost:8866/predict/ocr_system \ No newline at end of file diff --git a/doc/doc_ch/algorithm_overview.md b/doc/doc_ch/algorithm_overview.md index c4a3b3255f7367aec672387272d47b64a02658ee..440b392227182f50396f0e66ca8250bc6bfc1c0d 100644 --- a/doc/doc_ch/algorithm_overview.md +++ b/doc/doc_ch/algorithm_overview.md @@ -17,17 +17,17 @@ PaddleOCR开源的文本检测算法列表： |模型|骨干网络|precision|recall|Hmean|下载链接| |-|-|-|-|-|-| -|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[下载链接](https://paddleocr.bj.bcebos.com/det_r50_vd_east.tar)| -|EAST|MobileNetV3|81.67%|79.83%|80.74%|[下载链接](https://paddleocr.bj.bcebos.com/det_mv3_east.tar)| -|DB|ResNet50_vd|83.79%|80.65%|82.19%|[下载链接](https://paddleocr.bj.bcebos.com/det_r50_vd_db.tar)| -|DB|MobileNetV3|75.92%|73.18%|74.53%|[下载链接](https://paddleocr.bj.bcebos.com/det_mv3_db.tar)| -|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[下载链接](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_icdar2015.tar)| +|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[下载链接](link)| +|EAST|MobileNetV3|81.67%|79.83%|80.74%|[下载链接](link)| +|DB|ResNet50_vd|83.79%|80.65%|82.19%|[下载链接](link)| +|DB|MobileNetV3|75.92%|73.18%|74.53%|[下载链接](link)| +|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[下载链接](link))| 在Total-text文本检测公开数据集上，算法效果如下： |模型|骨干网络|precision|recall|Hmean|下载链接| |-|-|-|-|-|-| -|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[下载链接](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar)| +|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[下载链接](link)| **说明：** SAST模型训练额外加入了icdar2013、icdar2017、COCO-Text、ArT等公开数据集进行调优。PaddleOCR用到的经过整理格式的英文公开数据集下载：[百度云地址](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (提取码: 2bpi) @@ -37,28 +37,23 @@ PaddleOCR文本检测算法的训练和使用请参考文档教程中[模型训 ### 2.文本识别算法 -PaddleOCR开源的文本识别算法列表： +PaddleOCR基于动态图开源的文本识别算法列表： - [x] CRNN([paper](https://arxiv.org/abs/1507.05717))（ppocr推荐） - [x] Rosetta([paper](https://arxiv.org/abs/1910.05085)) - [x] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) -- [x] RARE([paper](https://arxiv.org/abs/1603.03915v1)) -- [x] SRN([paper](https://arxiv.org/abs/2003.12294)) +- [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon +- [ ] SRN([paper](https://arxiv.org/abs/2003.12294)) coming soon 参考[DTRB](https://arxiv.org/abs/1904.01906)文字识别训练和评估流程，使用MJSynth和SynthText两个文字识别数据集训练，在IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE数据集上进行评估，算法效果如下： |模型|骨干网络|Avg Accuracy|模型存储命名|下载链接| |-|-|-|-|-| -|Rosetta|Resnet34_vd|80.24%|rec_r34_vd_none_none_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_none_ctc.tar)| -|Rosetta|MobileNetV3|78.16%|rec_mv3_none_none_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_mv3_none_none_ctc.tar)| -|CRNN|Resnet34_vd|82.20%|rec_r34_vd_none_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_r34_vd_none_bilstm_ctc.tar)| -|CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar)| -|STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar)| -|STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[下载链接](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_ctc.tar)| -|RARE|Resnet34_vd|84.90%|rec_r34_vd_tps_bilstm_attn|[下载链接](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_attn.tar)| -|RARE|MobileNetV3|83.32%|rec_mv3_tps_bilstm_attn|[下载链接](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_attn.tar)| -|SRN|Resnet50_vd_fpn|88.33%|rec_r50fpn_vd_none_srn|[下载链接](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar)| - -**说明：** SRN模型使用了数据扰动方法对上述提到对两个训练集进行增广，增广后的数据可以在[百度网盘](https://pan.baidu.com/s/1-HSZ-ZVdqBF2HaBZ5pRAKA)上下载，提取码: y3ry。 -原始论文使用两阶段训练平均精度为89.74%，PaddleOCR中使用one-stage训练，平均精度为88.33%。两种预训练权重均在[下载链接](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar)中。 +|Rosetta|Resnet34_vd|80.24%|rec_r34_vd_none_none_ctc|[下载链接](link)| +|Rosetta|MobileNetV3|78.16%|rec_mv3_none_none_ctc|[下载链接](link)| +|CRNN|Resnet34_vd|82.20%|rec_r34_vd_none_bilstm_ctc|[下载链接](link)| +|CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[下载链接](link)| +|STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[下载链接](link)| +|STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[下载链接](link)| + PaddleOCR文本识别算法的训练和使用请参考文档教程中[模型训练/评估中的文本识别部分](./recognition.md)。 diff --git a/doc/doc_ch/config.md b/doc/doc_ch/config.md index 2cc502cadf5101c4321ca7543647dd90ea7e0466..92cb4ed6994d186a0591f4552150b8e8d31c7f15 100644 --- a/doc/doc_ch/config.md +++ b/doc/doc_ch/config.md @@ -10,14 +10,14 @@ ## 配置文件参数介绍 -以 `rec_chinese_lite_train_v1.1.yml ` 为例 -### Global +以 `rec_chinese_lite_train_v2.0.yml ` 为例 +### Global | 字段 | 用途 | 默认值 | 备注 | | :----------------------: | :---------------------: | :--------------: | :--------------------: | | use_gpu | 设置代码是否在gpu运行 | true | \ | | epoch_num | 最大训练epoch数 | 500 | \ | -| log_smooth_window | 滑动窗口大小 | 20 | \ | +| log_smooth_window | log队列长度，每次打印输出队列里的中间值 | 20 | \ | | print_batch_step | 设置打印log间隔 | 10 | \ | | save_model_dir | 设置模型保存路径 | output/{算法名称} | \ | | save_epoch_step | 设置模型保存间隔 | 3 | \ | @@ -119,4 +119,4 @@ | shuffle | 每个epoch是否将数据集顺序打乱 | True | \ | | batch_size_per_card | 训练时单卡batch size | 256 | \ | | drop_last | 是否丢弃因数据集样本数不能被 batch_size 整除而产生的最后一个不完整的mini-batch | True | \ | -| num_workers | 用于加载数据的子进程个数，若为0即为不开启子进程，在主进程中进行数据加载 | 8 | \ | \ No newline at end of file +| num_workers | 用于加载数据的子进程个数，若为0即为不开启子进程，在主进程中进行数据加载 | 8 | \ | diff --git a/doc/doc_ch/inference.md b/doc/doc_ch/inference.md index 0432695a3fbb31f04122c4134490ff465e445b4c..663533c492ab5dc0bd22cc79bd95c9d1d194d854 100644 --- a/doc/doc_ch/inference.md +++ b/doc/doc_ch/inference.md @@ -1,11 +1,11 @@ # 基于Python预测引擎推理 -inference 模型（`fluid.io.save_inference_model`保存的模型） +inference 模型（`paddle.jit.save`保存的模型） 一般是模型训练完成后保存的固化模型，多用于预测部署。训练过程中保存的模型是checkpoints模型，保存的是模型的参数，多用于恢复训练等。 -与checkpoints模型相比，inference 模型会额外保存模型的结构信息，在预测部署、加速推理上性能优越，灵活方便，适合与实际系统集成。更详细的介绍请参考文档[分类预测框架](https://github.com/PaddlePaddle/PaddleClas/blob/master/docs/zh_CN/extension/paddle_inference.md). +与checkpoints模型相比，inference 模型会额外保存模型的结构信息，在预测部署、加速推理上性能优越，灵活方便，适合与实际系统集成。 -接下来首先介绍如何将训练的模型转换成inference模型，然后将依次介绍文本检测、文本识别以及两者串联基于预测引擎推理。 +接下来首先介绍如何将训练的模型转换成inference模型，然后将依次介绍文本检测、文本角度分类器、文本识别以及三者串联基于预测引擎推理。 - [一、训练模型转inference模型](#训练模型转inference模型) @@ -23,9 +23,8 @@ inference 模型（`fluid.io.save_inference_model`保存的模型） - [1. 超轻量中文识别模型推理](#超轻量中文识别模型推理) - [2. 基于CTC损失的识别模型推理](#基于CTC损失的识别模型推理) - [3. 基于Attention损失的识别模型推理](#基于Attention损失的识别模型推理) - - [4. 基于SRN损失的识别模型推理](#基于SRN损失的识别模型推理) - - [5. 自定义文本识别字典的推理](#自定义文本识别字典的推理) - - [6. 多语言模型的推理](#多语言模型的推理) + - [4. 自定义文本识别字典的推理](#自定义文本识别字典的推理) + - [5. 多语言模型的推理](#多语言模型的推理) - [四、方向分类模型推理](#方向识别模型推理) - [1. 方向分类模型推理](#方向分类模型推理) @@ -42,24 +41,25 @@ inference 模型（`fluid.io.save_inference_model`保存的模型） 下载超轻量级中文检测模型： ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v1.1_det_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_det_train.tar -C ./ch_lite/ ``` 上述模型是以MobileNetV3为backbone训练的DB算法，将训练好的模型转换成inference模型只需要运行如下命令： ``` -# -c后面设置训练算法的yml配置文件 -# -o配置可选参数 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 +# -c 后面设置训练算法的yml配置文件 +# -o 配置可选参数 +# Global.pretrained_model 参数设置待转换的训练模型地址，不用添加文件后缀 .pdmodel，.pdopt或.pdparams。 +# Global.load_static_weights 参数需要设置为 False。 # Global.save_inference_dir参数设置转换的模型将保存的地址。 -python3 tools/export_model.py -c configs/det/det_mv3_db_v1.1.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_det_train/best_accuracy Global.save_inference_dir=./inference/det_db/ +python3 tools/export_model.py -c configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_det_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_db/ ``` -转inference模型时，使用的配置文件和训练时使用的配置文件相同。另外，还需要设置配置文件中的`Global.checkpoints`、`Global.save_inference_dir`参数。 -其中`Global.checkpoints`指向训练中保存的模型参数文件，`Global.save_inference_dir`是生成的inference模型要保存的目录。 -转换成功后，在`save_inference_dir`目录下有两个文件： +转inference模型时，使用的配置文件和训练时使用的配置文件相同。另外，还需要设置配置文件中的`Global.pretrained_model`参数，其指向训练中保存的模型参数文件。 +转换成功后，在模型保存目录下有三个文件： ``` inference/det_db/ - └─ model 检测inference模型的program文件 - └─ params 检测inference模型的参数文件 + ├── inference.pdiparams # 检测inference模型的参数文件 + ├── inference.pdiparams.info # 检测inference模型的参数信息，可忽略 + └── inference.pdmodel # 检测inference模型的program文件 ``` @@ -67,27 +67,28 @@ inference/det_db/ 下载超轻量中文识别模型： ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v1.1_rec_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_rec_train.tar -C ./ch_lite/ ``` 识别模型转inference模型与检测的方式相同，如下： ``` -# -c后面设置训练算法的yml配置文件 -# -o配置可选参数 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 +# -c 后面设置训练算法的yml配置文件 +# -o 配置可选参数 +# Global.pretrained_model 参数设置待转换的训练模型地址，不用添加文件后缀 .pdmodel，.pdopt或.pdparams。 +# Global.load_static_weights 参数需要设置为 False。 # Global.save_inference_dir参数设置转换的模型将保存的地址。 -python3 tools/export_model.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_rec_train/best_accuracy \ - Global.save_inference_dir=./inference/rec_crnn/ +python3 tools/export_model.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_rec_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/rec_crnn/ ``` **注意：**如果您是在自己的数据集上训练的模型，并且调整了中文字符的字典文件，请注意修改配置文件中的`character_dict_path`是否是所需要的字典文件。 -转换成功后，在目录下有两个文件： +转换成功后，在目录下有三个文件： ``` /inference/rec_crnn/ - └─ model 识别inference模型的program文件 - └─ params 识别inference模型的参数文件 + ├── inference.pdiparams # 识别inference模型的参数文件 + ├── inference.pdiparams.info # 识别inference模型的参数信息，可忽略 + └── inference.pdmodel # 识别inference模型的program文件 ``` @@ -95,25 +96,26 @@ python3 tools/export_model.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_trai 下载方向分类模型： ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v1.1_cls_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_cls_train.tar -C ./ch_lite/ ``` 方向分类模型转inference模型与检测的方式相同，如下： ``` -# -c后面设置训练算法的yml配置文件 -# -o配置可选参数 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 +# -c 后面设置训练算法的yml配置文件 +# -o 配置可选参数 +# Global.pretrained_model 参数设置待转换的训练模型地址，不用添加文件后缀 .pdmodel，.pdopt或.pdparams。 +# Global.load_static_weights 参数需要设置为 False。 # Global.save_inference_dir参数设置转换的模型将保存的地址。 -python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_cls_train/best_accuracy \ - Global.save_inference_dir=./inference/cls/ +python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_cls_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/cls/ ``` -转换成功后，在目录下有两个文件： +转换成功后，在目录下有三个文件： ``` /inference/cls/ - └─ model 识别inference模型的program文件 - └─ params 识别inference模型的参数文件 + ├── inference.pdiparams # 分类inference模型的参数文件 + ├── inference.pdiparams.info # 分类inference模型的参数信息，可忽略 + └── inference.pdmodel # 分类inference模型的program文件 ``` @@ -134,10 +136,12 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_di ![](../imgs_results/det_res_2.jpg) -通过设置参数`det_max_side_len`的大小，改变检测算法中图片规范化的最大值。当图片的长宽都小于`det_max_side_len`，则使用原图预测，否则将图片等比例缩放到最大值，进行预测。该参数默认设置为`det_max_side_len=960`。 如果输入图片的分辨率比较大，而且想使用更大的分辨率预测，可以执行如下命令： +通过参数`limit_type`和`det_limit_side_len`来对图片的尺寸进行限制限，`limit_type=max`为限制长边长度<`det_limit_side_len`，`limit_type=min`为限制短边长度>`det_limit_side_len`, +图片不满足限制条件时(`limit_type=max`时长边长度>`det_limit_side_len`或`limit_type=min`时短边长度<`det_limit_side_len`)，将对图片进行等比例缩放。 +该参数默认设置为`limit_type='max',det_max_side_len=960`。 如果输入图片的分辨率比较大，而且想使用更大的分辨率预测，可以执行如下命令： ``` -python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_max_side_len=1200 +python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_limit_type=max --det_limit_side_len=1200 ``` 如果想使用CPU进行预测，执行命令如下 @@ -148,14 +152,10 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_di ### 2. DB文本检测模型推理 -首先将DB文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/det_r50_vd_db.tar))，可以使用如下命令进行转换： +首先将DB文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（ [模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_db_v2.0_train.tar) )，可以使用如下命令进行转换： ``` -# -c后面设置训练算法的yml配置文件 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 -# Global.save_inference_dir参数设置转换的模型将保存的地址。 - -python3 tools/export_model.py -c configs/det/det_r50_vd_db.yml -o Global.checkpoints="./models/det_r50_vd_db/best_accuracy" Global.save_inference_dir="./inference/det_db" +python3 tools/export_model.py -c configs/det/det_r50_vd_db.yml -o Global.pretrained_model=./det_r50_vd_db_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_db ``` DB文本检测模型推理，可以执行如下命令： @@ -173,14 +173,10 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs_en/img_10.jpg" --det_ ### 3. EAST文本检测模型推理 -首先将EAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/det_r50_vd_east.tar))，可以使用如下命令进行转换： +首先将EAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例（ [模型下载地址 (coming soon)](link) )，可以使用如下命令进行转换： ``` -# -c后面设置训练算法的yml配置文件 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 -# Global.save_inference_dir参数设置转换的模型将保存的地址。 - -python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.checkpoints="./models/det_r50_vd_east/best_accuracy" Global.save_inference_dir="./inference/det_east" +python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.pretrained_model=./det_r50_vd_east_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_east ``` **EAST文本检测模型推理，需要设置参数`--det_algorithm="EAST"`**，可以执行如下命令： @@ -190,7 +186,7 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img ``` 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： -![](../imgs_results/det_res_img_10_east.jpg) +(coming soon) **注意**：本代码库中，EAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。 @@ -198,9 +194,10 @@ python3 tools/infer/predict_det.py --det_algorithm="EAST" --image_dir="./doc/img ### 4. SAST文本检测模型推理 #### (1). 四边形文本检测模型（ICDAR2015） -首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例([模型下载地址](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_icdar2015.tar))，可以使用如下命令进行转换： +首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例([模型下载地址(coming soon)](link))，可以使用如下命令进行转换： ``` -python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.checkpoints="./models/sast_r50_vd_icdar2015/best_accuracy" Global.save_inference_dir="./inference/det_sast_ic15" +python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.pretrained_model=./det_r50_vd_sast_icdar15_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_ic15 + ``` **SAST文本检测模型推理，需要设置参数`--det_algorithm="SAST"`**，可以执行如下命令： ``` @@ -208,13 +205,14 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img ``` 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： -![](../imgs_results/det_res_img_10_sast.jpg) +(coming soon) #### (2). 弯曲文本检测模型（Total-Text） -首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在Total-Text英文数据集训练的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar))，可以使用如下命令进行转换： +首先将SAST文本检测训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在Total-Text英文数据集训练的模型为例（[模型下载地址(coming soon)](link))，可以使用如下命令进行转换： ``` -python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.checkpoints="./models/sast_r50_vd_total_text/best_accuracy" Global.save_inference_dir="./inference/det_sast_tt" +python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.pretrained_model=./det_r50_vd_sast_totaltext_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_tt + ``` **SAST文本检测模型推理，需要设置参数`--det_algorithm="SAST"`，同时，还需要增加参数`--det_sast_polygon=True`，**可以执行如下命令： @@ -223,7 +221,7 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img ``` 可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'det_res'。结果示例如下： -![](../imgs_results/det_res_img623_sast.jpg) +(coming soon) **注意**：本代码库中，SAST后处理Locality-Aware NMS有python和c++两种版本，c++版速度明显快于python版。由于c++版本nms编译版本问题，只有python3.5环境下会调用c++版nms，其他情况将调用python版nms。 @@ -247,46 +245,46 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/ch/word_4.jpg" 执行命令后，上面图像的预测结果（识别的文本和得分）会打印到屏幕上，示例如下： -Predicts of ./doc/imgs_words/ch/word_4.jpg:['实力活力', 0.89552695] - +```bash +Predicts of ./doc/imgs_words/ch/word_4.jpg:('实力活力', 0.98458153) +``` ### 2. 基于CTC损失的识别模型推理 -我们以STAR-Net为例，介绍基于CTC损失的识别模型推理。 CRNN和Rosetta使用方式类似，不用设置识别算法参数rec_algorithm。 +我们以 CRNN 为例，介绍基于CTC损失的识别模型推理。 Rosetta 使用方式类似，不用设置识别算法参数rec_algorithm。 -首先将STAR-Net文本识别训练过程中保存的模型，转换成inference model。以基于Resnet34_vd骨干网络，使用MJSynth和SynthText两个英文文本识别合成数据集训练 -的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar))，可以使用如下命令进行转换： +首先将 CRNN 文本识别训练过程中保存的模型，转换成inference model。以基于Resnet34_vd骨干网络，使用MJSynth和SynthText两个英文文本识别合成数据集训练 +的模型为例（ [模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_bilstm_ctc_v2.0_train.tar) )，可以使用如下命令进行转换： ``` -# -c后面设置训练算法的yml配置文件 -# Global.checkpoints参数设置待转换的训练模型地址，不用添加文件后缀.pdmodel，.pdopt或.pdparams。 -# Global.save_inference_dir参数设置转换的模型将保存的地址。 +python3 tools/export_model.py -c configs/rec/rec_r34_vd_none_bilstm_ctc.yml -o Global.pretrained_model=./rec_r34_vd_none_bilstm_ctc_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/rec_crnn -python3 tools/export_model.py -c configs/rec/rec_r34_vd_tps_bilstm_ctc.yml -o Global.checkpoints="./models/rec_r34_vd_tps_bilstm_ctc/best_accuracy" Global.save_inference_dir="./inference/starnet" ``` -STAR-Net文本识别模型推理，可以执行如下命令： +CRNN 文本识别模型推理，可以执行如下命令： ``` -python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/starnet/" --rec_image_shape="3, 32, 100" --rec_char_type="en" +python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rec_crnn/" --rec_image_shape="3, 32, 100" --rec_char_type="en" ``` ### 3. 基于Attention损失的识别模型推理 基于Attention损失的识别模型与ctc不同，需要额外设置识别算法参数 --rec_algorithm="RARE" - RARE 文本识别模型推理，可以执行如下命令： ``` python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rare/" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_algorithm="RARE" + ``` ![](../imgs_words_en/word_336.png) 执行命令后，上面图像的识别结果如下： -Predicts of ./doc/imgs_words_en/word_336.png:['super', 0.9999555] +```bash +Predicts of ./doc/imgs_words_en/word_336.png:('super', 0.9999073) +``` **注意**：由于上述模型是参考[DTRB](https://arxiv.org/abs/1904.01906)文本识别训练和评估流程，与超轻量级中文识别模型训练有两方面不同： @@ -298,29 +296,16 @@ Predicts of ./doc/imgs_words_en/word_336.png:['super', 0.9999555] self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz" dict_character = list(self.character_str) ``` - -### 4. 基于SRN损失的识别模型推理 -基于SRN损失的识别模型，需要额外设置识别算法参数 --rec_algorithm="SRN"。 同时需要保证预测shape与训练时一致，如： --rec_image_shape="1, 64, 256" +### 4. 自定义文本识别字典的推理 +如果训练时修改了文本的字典，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径，并且设置 `rec_char_type=ch` ``` -python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" \ - --rec_model_dir="./inference/srn/" \ - --rec_image_shape="1, 64, 256" \ - --rec_char_type="en" \ - --rec_algorithm="SRN" -``` - - -### 5. 自定义文本识别字典的推理 -如果训练时修改了文本的字典，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径 - -``` -python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_char_dict_path="your text dict path" +python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="ch" --rec_char_dict_path="your text dict path" ``` -### 6. 多语言模型的推理 +### 5. 多语言模型的推理 如果您需要预测的是其他语言模型，在使用inference模型预测时，需要通过`--rec_char_dict_path`指定使用的字典路径, 同时为了得到正确的可视化结果， 需要通过 `--vis_font_path` 指定可视化的字体路径，`doc/` 路径下有默认提供的小语种字体，例如韩文识别： @@ -331,9 +316,7 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/korean/1.jpg" - 执行命令后，上图的预测结果为： ``` text -2020-09-19 16:15:05,076-INFO: index: [205 206 38 39] -2020-09-19 16:15:05,077-INFO: word : 바탕으로 -2020-09-19 16:15:05,077-INFO: score: 0.9171358942985535 +Predicts of ./doc/imgs_words/korean/1.jpg:('바탕으로', 0.9948904) ``` @@ -350,11 +333,13 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/korean/1.jpg" - python3 tools/infer/predict_cls.py --image_dir="./doc/imgs_words/ch/word_4.jpg" --cls_model_dir="./inference/cls/" ``` -![](../imgs_words/ch/word_4.jpg) +![](../imgs_words/ch/word_1.jpg) 执行命令后，上面图像的预测结果（分类的方向和得分）会打印到屏幕上，示例如下： -Predicts of ./doc/imgs_words/ch/word_4.jpg:['0', 0.9999963] +``` +Predicts of ./doc/imgs_words/ch/word_4.jpg:['0', 0.9999982] +``` ## 五、文本检测、方向分类和文字识别串联推理 @@ -394,4 +379,4 @@ python3 tools/infer/predict_system.py --image_dir="./doc/imgs_en/img_10.jpg" --d 执行命令后，识别结果图像如下： -![](../imgs_results/img_10.jpg) +(coming soon) diff --git a/doc/doc_ch/models_list.md b/doc/doc_ch/models_list.md index c85def4efbc8abdd5eeadbd7a4641b5776694491..b281e1e736f6c3747c2ae07188dc6f87abfc67a8 100644 --- a/doc/doc_ch/models_list.md +++ b/doc/doc_ch/models_list.md @@ -1,4 +1,4 @@ -## OCR模型列表（V1.1，9月22日更新） +## OCR模型列表（V2.0，2020年12月12日更新） - [一、文本检测模型](#文本检测模型) - [二、文本识别模型](#文本识别模型) @@ -10,19 +10,20 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训练模型`、`slim模型`，模型区别说明如下： |模型类型|模型格式|简介| -|-|-|-| -|推理模型|model、params|用于python预测引擎推理，[详情](./inference.md)| -|训练模型、预训练模型|\*.pdmodel、\*.pdopt、\*.pdparams|训练过程中保存的checkpoints模型，保存的是模型的参数，多用于模型指标评估和恢复训练| +|--- | --- | --- | +|推理模型|inference.pdmodel、inference.pdiparams|用于python预测引擎推理，[详情](./inference.md)| +|训练模型、预训练模型|\*.pdparams、\*.pdopt、\*.states |训练过程中保存的模型的参数、优化器状态和训练中间信息，多用于模型指标评估和恢复训练| |slim模型|\*.nb|用于lite部署| ### 一、文本检测模型 + |模型名称|模型简介|配置文件|推理模型大小|下载地址| -|-|-|-|-|-| -|ch_ppocr_mobile_slim_v1.1_det|slim裁剪版超轻量模型，支持中英文、多语种文本检测|[det_mv3_db_v1.1.yml](../../configs/det/det_mv3_db_v1.1.yml)|1.4M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/det/ch_ppocr_mobile_v1.1_det_prune_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_det_prune_opt.nb)| -|ch_ppocr_mobile_v1.1_det|原始超轻量模型，支持中英文、多语种文本检测|[det_mv3_db_v1.1.yml](../../configs/det/det_mv3_db_v1.1.yml)|2.6M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar)| -|ch_ppocr_server_v1.1_det|通用模型，支持中英文、多语种文本检测，比超轻量模型更大，但效果更好|[det_r18_vd_db_v1.1.yml](../../configs/det/det_r18_vd_db_v1.1.yml)|47.2M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_train.tar)| +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_det|slim裁剪版超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)| |[推理模型 (coming soon)](link) / [slim模型 (coming soon)](link)| +|ch_ppocr_mobile_v2.0_det|原始超轻量模型，支持中英文、多语种文本检测|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)|3M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)| +|ch_ppocr_server_v2.0_det|通用模型，支持中英文、多语种文本检测，比超轻量模型更大，但效果更好|[ch_det_res18_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml)|47M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)| @@ -30,42 +31,44 @@ PaddleOCR提供的可下载模型包括`推理模型`、`训练模型`、`预训 #### 1. 中文识别模型 + |模型名称|模型简介|配置文件|推理模型大小|下载地址| -|-|-|-|-|-| -|ch_ppocr_mobile_slim_v1.1_rec|slim裁剪量化版超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml)|1.6M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/rec/ch_ppocr_mobile_v1.1_rec_quant_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_rec_quant_opt.nb) | -|ch_ppocr_mobile_v1.1_rec|原始超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml)|4.6M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_pre.tar) | -|ch_ppocr_server_v1.1_rec|通用模型，支持中英文、数字识别|[rec_chinese_common_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yml)|105M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_pre.tar) | +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)| |[推理模型 (coming soon)](link) / [slim模型 (coming soon)](link) | +|ch_ppocr_mobile_v2.0_rec|原始超轻量模型，支持中英文、数字识别|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)|3.71M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +|ch_ppocr_server_v2.0_rec|通用模型，支持中英文、数字识别|[rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml)|94.8M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | **说明：** `训练模型`是基于预训练模型在真实数据与竖排合成文本数据上finetune得到的模型，在真实应用场景中有着更好的表现，`预训练模型`则是直接基于全量真实数据与合成数据训练得到，更适合用于在自己的数据集上finetune。 #### 2. 英文识别模型 + |模型名称|模型简介|配置文件|推理模型大小|下载地址| -|-|-|-|-|-| -|en_ppocr_mobile_slim_v1.1_rec|slim裁剪量化版超轻量模型，支持英文、数字识别|[rec_en_lite_train.yml](../../configs/rec/multi_languages/rec_en_lite_train.yml)|0.9M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/en/en_ppocr_mobile_v1.1_rec_quant_infer.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/en/en_ppocr_mobile_v1.1_rec_quant_opt.nb) | -|en_ppocr_mobile_v1.1_rec|原始超轻量模型，支持英文、数字识别|[rec_en_lite_train.yml](../../configs/rec/multi_languages/rec_en_lite_train.yml)|2.0M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/en/en_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/en/en_ppocr_mobile_v1.1_rec_train.tar) | +| --- | --- | --- | --- | --- | +|en_number_mobile_slim_v2.0_rec|slim裁剪量化版超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)| |[推理模型 (coming soon )](link) / [slim模型 (coming soon)](link) | +|en_number_mobile_v2.0_rec|原始超轻量模型，支持英文、数字识别|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)|2.56M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_train.tar) | #### 3. 多语言识别模型（更多语言持续更新中...） + |模型名称|模型简介|配置文件|推理模型大小|下载地址| -|-|-|-|-|-| -| french_ppocr_mobile_v1.1_rec |法文识别|[rec_french_lite_train.yml](../../configs/rec/multi_languages/rec_french_lite_train.yml)|2.1M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_train.tar) | -| german_ppocr_mobile_v1.1_rec |德文识别|[rec_ger_lite_train.yml](../../configs/rec/multi_languages/rec_ger_lite_train.yml)|2.1M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_train.tar) | -| korean_ppocr_mobile_v1.1_rec |韩文识别|[rec_korean_lite_train.yml](../../configs/rec/multi_languages/rec_korean_lite_train.yml)|3.4M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_train.tar) | -| japan_ppocr_mobile_v1.1_rec |日文识别|[rec_japan_lite_train.yml](../../configs/rec/multi_languages/rec_japan_lite_train.yml)|3.7M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_train.tar) | +| --- | --- | --- | --- | --- | +| french_mobile_v2.0_rec |法文识别|[rec_french_lite_train.yml](../../configs/rec/multi_language/rec_french_lite_train.yml)|2.65M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_train.tar) | +| german_mobile_v2.0_rec |德文识别|[rec_german_lite_train.yml](../../configs/rec/multi_language/rec_german_lite_train.yml)|2.65M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_train.tar) | +| korean_mobile_v2.0_rec |韩文识别|[rec_korean_lite_train.yml](../../configs/rec/multi_language/rec_korean_lite_train.yml)|3.9M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_train.tar) | +| japan_mobile_v2.0_rec |日文识别|[rec_japan_lite_train.yml](../../configs/rec/multi_language/rec_japan_lite_train.yml)|4.23M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_train.tar) | ### 三、文本方向分类模型 + |模型名称|模型简介|配置文件|推理模型大小|下载地址| -|-|-|-|-|-| -|ch_ppocr_mobile_v1.1_cls_quant|slim量化版模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|0.5M|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_quant_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_quant_train.tar) / [slim模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_cls_quant_opt.nb) | -|ch_ppocr_mobile_v1.1_cls|原始模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|850kb|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) | +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_cls|slim量化版模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)| |[推理模型 (coming soon)](link) / [训练模型](link) / [slim模型](link) | +|ch_ppocr_mobile_v2.0_cls|原始模型|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|1.38M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) | + +## OCR模型列表（V1.1，2020年9月22日更新） -## OCR模型列表（V1.0，7月16日更新） +[1.1系列模型地址](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/models_list.md) -|模型名称|模型简介|检测模型地址|识别模型地址|支持空格的识别模型地址| -|-|-|-|-|-| -|chinese_db_crnn_mobile|8.6M超轻量级中文OCR模型|[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db.tar) |[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn.tar) |[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance.tar) -|chinese_db_crnn_server|通用中文OCR模型|[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db.tar) |[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn.tar) |[推理模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance.tar) diff --git a/doc/doc_ch/quickstart.md b/doc/doc_ch/quickstart.md index 97c3d41dae337195f0ac1517a4c276a886e7cc94..eabf1d91cfcc5afad3b9495f63cd6379562342b9 100644 --- a/doc/doc_ch/quickstart.md +++ b/doc/doc_ch/quickstart.md @@ -5,16 +5,16 @@ 请先参考[快速安装](./installation.md)配置PaddleOCR运行环境。 -*注意：也可以通过 whl 包安装使用PaddleOCR，具体参考[Paddleocr Package使用说明](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_ch/whl.md)。* +*注意：也可以通过 whl 包安装使用PaddleOCR，具体参考[Paddleocr Package使用说明](./whl.md)。* ## 2.inference模型下载 -* 移动端和服务器端的检测与识别模型如下，更多模型下载（包括多语言），可以参考[PP-OCR v1.1 系列模型下载](../doc_ch/models_list.md) +* 移动端和服务器端的检测与识别模型如下，更多模型下载（包括多语言），可以参考[PP-OCR v2.0 系列模型下载](../doc_ch/models_list.md) | 模型简介 | 模型名称 |推荐场景 | 检测模型 | 方向分类器 | 识别模型 | | ------------ | --------------- | ----------------|---- | ---------- | -------- | -| 中英文超轻量OCR模型（8.1M） | ch_ppocr_mobile_v1.1_xx |移动端&服务器端|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar)|[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_pre.tar) | -| 中英文通用OCR模型（155.1M） |ch_ppocr_server_v1.1_xx|服务器端 |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_pre.tar) | +| 中英文超轻量OCR模型（8.1M） | ch_ppocr_mobile_v2.0_xx |移动端&服务器端|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +| 中英文通用OCR模型（143M） | ch_ppocr_server_v2.0_xx |服务器端 |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [预训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | * windows 环境下如果没有安装wget,下载模型时可将链接复制到浏览器中下载，并解压放置在相应目录下 @@ -37,44 +37,45 @@ cd .. ``` mkdir inference && cd inference # 下载超轻量级中文OCR模型的检测模型并解压 -wget https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar && tar xf ch_ppocr_mobile_v1.1_det_infer.tar +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_ppocr_mobile_v2.0_det_infer.tar # 下载超轻量级中文OCR模型的识别模型并解压 -wget https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar && tar xf ch_ppocr_mobile_v1.1_rec_infer.tar +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar # 下载超轻量级中文OCR模型的文本方向分类器模型并解压 -wget https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar && tar xf ch_ppocr_mobile_v1.1_cls_infer.tar +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar && tar xf ch_ppocr_mobile_v2.0_cls_infer.tar cd .. ``` 解压完毕后应有如下文件结构： ``` -|-inference - |-ch_ppocr_mobile_v1.1_det_infer - |- model - |- params - |-ch_ppocr_mobile_v1.1_rec_infer - |- model - |- params - |-ch_ppocr_mobile-v1.1_cls_infer - |- model - |- params - ... +├── ch_ppocr_mobile_v2.0_cls_infer +│ ├── inference.pdiparams +│ ├── inference.pdiparams.info +│ └── inference.pdmodel +├── ch_ppocr_mobile_v2.0_det_infer +│ ├── inference.pdiparams +│ ├── inference.pdiparams.info +│ └── inference.pdmodel +├── ch_ppocr_mobile_v2.0_rec_infer + ├── inference.pdiparams + ├── inference.pdiparams.info + └── inference.pdmodel ``` ## 3.单张图像或者图像集合预测 -以下代码实现了文本检测、识别串联推理，在执行预测时，需要通过参数image_dir指定单张图像或者图像集合的路径、参数`det_model_dir`指定检测inference模型的路径、参数`rec_model_dir`指定识别inference模型的路径、参数`use_angle_cls`指定是否使用方向分类器、参数`cls_model_dir`指定方向分类器inference模型的路径、参数`use_space_char`指定是否预测空格字符。可视化识别结果默认保存到`./inference_results`文件夹里面。 +以下代码实现了文本检测、方向分类器和识别串联推理，在执行预测时，需要通过参数image_dir指定单张图像或者图像集合的路径、参数`det_model_dir`指定检测inference模型的路径、参数`rec_model_dir`指定识别inference模型的路径、参数`use_angle_cls`指定是否使用方向分类器、参数`cls_model_dir`指定方向分类器inference模型的路径、参数`use_space_char`指定是否预测空格字符。可视化识别结果默认保存到`./inference_results`文件夹里面。 ```bash # 预测image_dir指定的单张图像 -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True # 预测image_dir指定的图像集合 -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True # 如果想使用CPU进行预测，需设置use_gpu参数为False -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True --use_gpu=False +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True --use_gpu=False ``` - 通用中文OCR模型 @@ -83,7 +84,7 @@ python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_mode ```bash # 预测image_dir指定的单张图像 -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_server_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_server_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_server_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_server_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True ``` * 注意： diff --git a/doc/doc_ch/recognition.md b/doc/doc_ch/recognition.md index 71be1e89dc561630315337ef11c52289e0756c00..87d60c5504d28c3cae660ebfd3765bb6893f163e 100644 --- a/doc/doc_ch/recognition.md +++ b/doc/doc_ch/recognition.md @@ -37,8 +37,6 @@ ln -sf /train_data/dataset 若您本地没有数据集，可以在官网下载 [icdar2015](http://rrc.cvc.uab.es/?ch=4&com=downloads) 数据，用于快速验证。也可以参考[DTRB](https://github.com/clovaai/deep-text-recognition-benchmark#download-lmdb-dataset-for-traininig-and-evaluation-from-here)，下载 benchmark 所需的lmdb格式数据集。 -如果希望复现SRN的论文指标，需要下载离线[增广数据](https://pan.baidu.com/s/1-HSZ-ZVdqBF2HaBZ5pRAKA),提取码: y3ry。增广数据是由MJSynth和SynthText做旋转和扰动得到的。数据下载完成后请解压到 {your_path}/PaddleOCR/train_data/data_lmdb_release/training/ 路径下。 - * 使用自己数据集 @@ -65,7 +63,7 @@ wget -P ./train_data/ic15_data https://paddleocr.bj.bcebos.com/dataset/rec_gt_t wget -P ./train_data/ic15_data https://paddleocr.bj.bcebos.com/dataset/rec_gt_test.txt ``` -PaddleOCR 也提供了数据格式转换脚本，可以将官网 label 转换支持的数据格式。 数据转换工具在 `train_data/gen_label.py`, 这里以训练集为例： +PaddleOCR 也提供了数据格式转换脚本，可以将官网 label 转换支持的数据格式。 数据转换工具在 `ppocr/utils/gen_label.py`, 这里以训练集为例： ``` # 将官网下载的标签文件转换为 rec_gt_label.txt @@ -116,17 +114,19 @@ n word_dict.txt 每行有一个单字，将字符与数字索引映射在一起，“and” 将被映射成 [2 5 1] -`ppocr/utils/ppocr_keys_v1.txt` 是一个包含6623个字符的中文字典， +`ppocr/utils/ppocr_keys_v1.txt` 是一个包含6623个字符的中文字典 -`ppocr/utils/ic15_dict.txt` 是一个包含36个字符的英文字典， +`ppocr/utils/ic15_dict.txt` 是一个包含36个字符的英文字典 `ppocr/utils/dict/french_dict.txt` 是一个包含118个字符的法文字典 -`ppocr/utils/dict/japan_dict.txt` 是一个包含4399个字符的法文字典 +`ppocr/utils/dict/japan_dict.txt` 是一个包含4399个字符的日文字典 + +`ppocr/utils/dict/korean_dict.txt` 是一个包含3636个字符的韩文字典 -`ppocr/utils/dict/korean_dict.txt` 是一个包含3636个字符的法文字典 +`ppocr/utils/dict/german_dict.txt` 是一个包含131个字符的德文字典 -`ppocr/utils/dict/german_dict.txt` 是一个包含131个字符的法文字典 +`ppocr/utils/dict/en_dict.txt` 是一个包含63个字符的英文字典 您可以按需使用。 @@ -142,9 +142,8 @@ word_dict.txt 每行有一个单字，将字符与数字索引映射在一起， - 添加空格类别 -如果希望支持识别"空格"类别, 请将yml文件中的 `use_space_char` 字段设置为 `true`。 +如果希望支持识别"空格"类别, 请将yml文件中的 `use_space_char` 字段设置为 `True`。 -**注意：`use_space_char` 仅在 `character_type=ch` 时生效** ### 启动训练 @@ -156,10 +155,10 @@ PaddleOCR提供了训练脚本、评估脚本和预测脚本，本节将以 CRNN ``` cd PaddleOCR/ # 下载MobileNetV3的预训练模型 -wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar +wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_bilstm_ctc_v2.0_train.tar # 解压模型参数 cd pretrain_models -tar -xf rec_mv3_none_bilstm_ctc.tar && rm -rf rec_mv3_none_bilstm_ctc.tar +tar -xf rec_mv3_none_bilstm_ctc_v2.0_train.tar && rm -rf rec_mv3_none_bilstm_ctc_v2.0_train.tar ``` 开始训练: @@ -167,10 +166,9 @@ tar -xf rec_mv3_none_bilstm_ctc.tar && rm -rf rec_mv3_none_bilstm_ctc.tar *如果您安装的是cpu版本，请将配置文件中的 `use_gpu` 字段修改为false* ``` -# GPU训练 支持单卡，多卡训练，通过CUDA_VISIBLE_DEVICES指定卡号 -export CUDA_VISIBLE_DEVICES=0,1,2,3 +# GPU训练 支持单卡，多卡训练，通过--gpus参数指定卡号 # 训练icdar15英文数据 并将训练日志保存为 tain_rec.log -python3 tools/train.py -c configs/rec/rec_icdar15_train.yml 2>&1 | tee train_rec.log +python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/rec/rec_icdar15_train.yml ``` - 数据增强 @@ -195,8 +193,8 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t | 配置文件 | 算法名称 | backbone | trans | seq | pred | | :--------: | :-------: | :-------: | :-------: | :-----: | :-----: | -| [rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml) | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | -| [rec_chinese_common_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yml) | CRNN | ResNet34_vd | None | BiLSTM | ctc | +| [rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml) | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | +| [rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml) | CRNN | ResNet34_vd | None | BiLSTM | ctc | | rec_chinese_lite_train.yml | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | | rec_chinese_common_train.yml | CRNN | ResNet34_vd | None | BiLSTM | ctc | | rec_icdar15_train.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc | @@ -206,43 +204,71 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t | rec_mv3_tps_bilstm_attn.yml | RARE | Mobilenet_v3 large 0.5 | tps | BiLSTM | attention | | rec_r34_vd_none_bilstm_ctc.yml | CRNN | Resnet34_vd | None | BiLSTM | ctc | | rec_r34_vd_none_none_ctc.yml | Rosetta | Resnet34_vd | None | None | ctc | -| rec_r34_vd_tps_bilstm_attn.yml | RARE | Resnet34_vd | tps | BiLSTM | attention | | rec_r34_vd_tps_bilstm_ctc.yml | STARNet | Resnet34_vd | tps | BiLSTM | ctc | -| rec_r50fpn_vd_none_srn.yml | SRN | Resnet50_fpn_vd | None | rnn | srn | -训练中文数据，推荐使用[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件： +训练中文数据，推荐使用[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件： -以 `rec_mv3_none_none_ctc.yml` 为例： +以 `rec_chinese_lite_train_v2.0.yml` 为例： ``` Global: ... - # 修改 image_shape 以适应长文本 - image_shape: [3, 32, 320] - ... + # 添加自定义字典，如修改字典请将路径指向新字典 + character_dict_path: ppocr/utils/ppocr_keys_v1.txt # 修改字符类型 character_type: ch - # 添加自定义字典，如修改字典请将路径指向新字典 - character_dict_path: ./ppocr/utils/ppocr_keys_v1.txt - # 训练时添加数据增强 - distort: true - # 识别空格 - use_space_char: true - ... - # 修改reader类型 - reader_yml: ./configs/rec/rec_chinese_reader.yml ... + # 识别空格 + use_space_char: True -... Optimizer: ... # 添加学习率衰减策略 - decay: - function: cosine_decay - # 每个 epoch 包含 iter 数 - step_each_epoch: 20 - # 总共训练epoch数 - total_epoch: 1000 + lr: + name: Cosine + learning_rate: 0.001 + ... + +... + +Train: + dataset: + # 数据集格式，支持LMDBDateSet以及SimpleDataSet + name: SimpleDataSet + # 数据集路径 + data_dir: ./train_data/ + # 训练集标签文件 + label_file_list: ["./train_data/train_list.txt"] + transforms: + ... + - RecResizeImg: + # 修改 image_shape 以适应长文本 + image_shape: [3, 32, 320] + ... + loader: + ... + # 单卡训练的batch_size + batch_size_per_card: 256 + ... + +Eval: + dataset: + # 数据集格式，支持LMDBDateSet以及SimpleDataSet + name: SimpleDataSet + # 数据集路径 + data_dir: ./train_data + # 验证集标签文件 + label_file_list: ["./train_data/val_list.txt"] + transforms: + ... + - RecResizeImg: + # 修改 image_shape 以适应长文本 + image_shape: [3, 32, 320] + ... + loader: + # 单卡验证的batch_size + batch_size_per_card: 256 + ... ``` **注意，预测/评估时的配置文件请务必与训练一致。** @@ -270,39 +296,41 @@ Global: ... # 添加自定义字典，如修改字典请将路径指向新字典 character_dict_path: ./ppocr/utils/dict/french_dict.txt - # 训练时添加数据增强 - distort: true - # 识别空格 - use_space_char: true - ... - # 修改reader类型 - reader_yml: ./configs/rec/multi_languages/rec_french_reader.yml ... -... -``` - -同时需要修改数据读取文件 `rec_french_reader.yml`： - -``` -TrainReader: - ... - # 修改训练数据存放的目录名 - img_set_dir: ./train_data - # 修改 label 文件名称 - label_file_path: ./train_data/french_train.txt + # 识别空格 + use_space_char: True ... + +Train: + dataset: + # 数据集格式，支持LMDBDateSet以及SimpleDataSet + name: SimpleDataSet + # 数据集路径 + data_dir: ./train_data/ + # 训练集标签文件 + label_file_list: ["./train_data/french_train.txt"] + ... + +Eval: + dataset: + # 数据集格式，支持LMDBDateSet以及SimpleDataSet + name: SimpleDataSet + # 数据集路径 + data_dir: ./train_data + # 验证集标签文件 + label_file_list: ["./train_data/french_val.txt"] + ... ``` ### 评估 -评估数据集可以通过 `configs/rec/rec_icdar15_reader.yml` 修改EvalReader中的 `label_file_path` 设置。 +评估数据集可以通过 `configs/rec/rec_icdar15_train.yml` 修改Eval中的 `label_file_path` 设置。 *注意* 评估时必须确保配置文件中 infer_img 字段为空 ``` -export CUDA_VISIBLE_DEVICES=0 # GPU 评估， Global.checkpoints 为待测权重 -python3 tools/eval.py -c configs/rec/rec_icdar15_train.yml -o Global.checkpoints={path/to/weights}/best_accuracy +python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_icdar15_train.yml -o Global.checkpoints={path/to/weights}/best_accuracy ``` @@ -332,12 +360,12 @@ infer_img: doc/imgs_words/en/word_1.png word : joint ``` -预测使用的配置文件必须与训练一致，如您通过 `python3 tools/train.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml` 完成了中文模型的训练， +预测使用的配置文件必须与训练一致，如您通过 `python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml` 完成了中文模型的训练， 您可以使用如下命令进行中文模型预测。 ``` # 预测中文结果 -python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml -o Global.checkpoints={path/to/weights}/best_accuracy Global.infer_img=doc/imgs_words/ch/word_1.jpg +python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.checkpoints={path/to/weights}/best_accuracy Global.infer_img=doc/imgs_words/ch/word_1.jpg ``` 预测图片： diff --git a/doc/doc_ch/whl.md b/doc/doc_ch/whl.md index c51f32778f0e6e5069cb8d45e1632263b68569ad..587b443baf2ed92c1913b29f2dad45b812b44928 100644 --- a/doc/doc_ch/whl.md +++ b/doc/doc_ch/whl.md @@ -348,7 +348,7 @@ im_show.save('result.jpg') | cls_batch_num | 进行分类时，同时前向的图片数 |30 | | enable_mkldnn | 是否启用mkldnn | FALSE | | use_zero_copy_run | 是否通过zero_copy_run的方式进行前向 | FALSE | -| lang | 模型语言类型,目前支持 中文(ch)和英文(en) | ch | +| lang | 模型语言类型,目前支持 目前支持中英文(ch)、英文(en)、法语(french)、德语(german)、韩语(korean)、日语(japan) | ch | | det | 前向时使用启动检测 | TRUE | | rec | 前向时是否启动识别 | TRUE | | cls | 前向时是否启动分类 (命令行模式下使用use_angle_cls控制前向是否启动分类) | FALSE | diff --git a/doc/doc_en/algorithm_overview_en.md b/doc/doc_en/algorithm_overview_en.md index 2e21fd621971e062384a9323e79a8cf4498d7495..532ebd90cf149813acc9ad929840e1611766f652 100644 --- a/doc/doc_en/algorithm_overview_en.md +++ b/doc/doc_en/algorithm_overview_en.md @@ -19,17 +19,17 @@ On the ICDAR2015 dataset, the text detection result is as follows: |Model|Backbone|precision|recall|Hmean|Download link| |-|-|-|-|-|-| -|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[Download link](https://paddleocr.bj.bcebos.com/det_r50_vd_east.tar)| -|EAST|MobileNetV3|81.67%|79.83%|80.74%|[Download link](https://paddleocr.bj.bcebos.com/det_mv3_east.tar)| -|DB|ResNet50_vd|83.79%|80.65%|82.19%|[Download link](https://paddleocr.bj.bcebos.com/det_r50_vd_db.tar)| -|DB|MobileNetV3|75.92%|73.18%|74.53%|[Download link](https://paddleocr.bj.bcebos.com/det_mv3_db.tar)| -|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[Download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_icdar2015.tar)| +|EAST|ResNet50_vd|88.18%|85.51%|86.82%|[Download link](link)| +|EAST|MobileNetV3|81.67%|79.83%|80.74%|[Download link](link)| +|DB|ResNet50_vd|83.79%|80.65%|82.19%|[Download link](link)| +|DB|MobileNetV3|75.92%|73.18%|74.53%|[Download link](link)| +|SAST|ResNet50_vd|92.18%|82.96%|87.33%|[Download link](link)| On Total-Text dataset, the text detection result is as follows: |Model|Backbone|precision|recall|Hmean|Download link| |-|-|-|-|-|-| -|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[Download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar)| +|SAST|ResNet50_vd|88.74%|79.80%|84.03%|[Download link](link)| **Note：** Additional data, like icdar2013, icdar2017, COCO-Text, ArT, was added to the model training of SAST. Download English public dataset in organized format used by PaddleOCR from [Baidu Drive](https://pan.baidu.com/s/12cPnZcVuV1zn5DOd4mqjVw) (download code: 2bpi). @@ -42,8 +42,8 @@ PaddleOCR open-source text recognition algorithms list: - [x] CRNN([paper](https://arxiv.org/abs/1507.05717)) - [x] Rosetta([paper](https://arxiv.org/abs/1910.05085)) - [x] STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)) -- [x] RARE([paper](https://arxiv.org/abs/1603.03915v1)) -- [x] SRN([paper](https://arxiv.org/abs/2003.12294))(Baidu Self-Research) +- [ ] RARE([paper](https://arxiv.org/abs/1603.03915v1)) coming soon +- [ ] SRN([paper](https://arxiv.org/abs/2003.12294))(Baidu Self-Research) coming soon Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow: @@ -55,12 +55,6 @@ Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation r |CRNN|MobileNetV3|79.37%|rec_mv3_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar)| |STAR-Net|Resnet34_vd|83.93%|rec_r34_vd_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar)| |STAR-Net|MobileNetV3|81.56%|rec_mv3_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_ctc.tar)| -|RARE|Resnet34_vd|84.90%|rec_r34_vd_tps_bilstm_attn|[Download link](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_attn.tar)| -|RARE|MobileNetV3|83.32%|rec_mv3_tps_bilstm_attn|[Download link](https://paddleocr.bj.bcebos.com/rec_mv3_tps_bilstm_attn.tar)| -|SRN|Resnet50_vd_fpn|88.33%|rec_r50fpn_vd_none_srn|[Download link](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar)| -**Note：** SRN model uses data expansion method to expand the two training sets mentioned above, and the expanded data can be downloaded from [Baidu Drive](https://pan.baidu.com/s/1-HSZ-ZVdqBF2HaBZ5pRAKA) (download code: y3ry). - -The average accuracy of the two-stage training in the original paper is 89.74%, and that of one stage training in paddleocr is 88.33%. Both pre-trained weights can be downloaded [here](https://paddleocr.bj.bcebos.com/SRN/rec_r50fpn_vd_none_srn.tar). Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./doc/doc_en/recognition_en.md) diff --git a/doc/doc_en/config_en.md b/doc/doc_en/config_en.md index 574bb41b6b1735271f9c794b856c5efb32db424f..ada1678e4ed059bbd3af3a5fdee42afaed1fce01 100644 --- a/doc/doc_en/config_en.md +++ b/doc/doc_en/config_en.md @@ -9,14 +9,14 @@ The following list can be viewed through `--help` ## INTRODUCTION TO GLOBAL PARAMETERS OF CONFIGURATION FILE -Take rec_chinese_lite_train_v1.1.yml as an example -### Global +Take rec_chinese_lite_train_v2.0.yml as an example +### Global | Parameter | Use | Defaults | Note | | :----------------------: | :---------------------: | :--------------: | :--------------------: | | use_gpu | Set using GPU or not | true | \ | | epoch_num | Maximum training epoch number | 500 | \ | -| log_smooth_window | Sliding window size | 20 | \ | +| log_smooth_window | Log queue length, the median value in the queue each time will be printed | 20 | \ | | print_batch_step | Set print log interval | 10 | \ | | save_model_dir | Set model save path | output/{算法名称} | \ | | save_epoch_step | Set model save interval | 3 | \ | @@ -118,4 +118,4 @@ In ppocr, the network is divided into four stages: Transform, Backbone, Neck and | shuffle | Does each epoch disrupt the order of the data set | True | \ | | batch_size_per_card | Single card batch size during training | 256 | \ | | drop_last | Whether to discard the last incomplete mini-batch because the number of samples in the data set cannot be divisible by batch_size | True | \ | -| num_workers | The number of sub-processes used to load data, if it is 0, the sub-process is not started, and the data is loaded in the main process | 8 | \ | \ No newline at end of file +| num_workers | The number of sub-processes used to load data, if it is 0, the sub-process is not started, and the data is loaded in the main process | 8 | \ | diff --git a/doc/doc_en/inference_en.md b/doc/doc_en/inference_en.md index 609b65fa55a743acd72407a34288afc793885d3c..411a733dd062cf347d7a2e5d5d067739bda36819 100644 --- a/doc/doc_en/inference_en.md +++ b/doc/doc_en/inference_en.md @@ -1,13 +1,13 @@ # Reasoning based on Python prediction engine -The inference model (the model saved by `fluid.io.save_inference_model`) is generally a solidified model saved after the model training is completed, and is mostly used to give prediction in deployment. +The inference model (the model saved by `paddle.jit.save`) is generally a solidified model saved after the model training is completed, and is mostly used to give prediction in deployment. The model saved during the training process is the checkpoints model, which saves the parameters of the model and is mostly used to resume training. Compared with the checkpoints model, the inference model will additionally save the structural information of the model. It has superior performance in predicting in deployment and accelerating inferencing, is flexible and convenient, and is suitable for integration with actual systems. For more details, please refer to the document [Classification Framework](https://github.com/PaddlePaddle/PaddleClas/blob/master/docs/zh_CN/extension/paddle_inference.md). -Next, we first introduce how to convert a trained model into an inference model, and then we will introduce text detection, text recognition, and the concatenation of them based on inference model. +Next, we first introduce how to convert a trained model into an inference model, and then we will introduce text detection, text recognition, angle class, and the concatenation of them based on inference model. - [CONVERT TRAINING MODEL TO INFERENCE MODEL](#CONVERT) - [Convert detection model to inference model](#Convert_detection_model) @@ -26,9 +26,8 @@ Next, we first introduce how to convert a trained model into an inference model, - [1. LIGHTWEIGHT CHINESE MODEL](#LIGHTWEIGHT_RECOGNITION) - [2. CTC-BASED TEXT RECOGNITION MODEL INFERENCE](#CTC-BASED_RECOGNITION) - [3. ATTENTION-BASED TEXT RECOGNITION MODEL INFERENCE](#ATTENTION-BASED_RECOGNITION) - - [4. SRN-BASED TEXT RECOGNITION MODEL INFERENCE](#SRN-BASED_RECOGNITION) - - [5. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY](#USING_CUSTOM_CHARACTERS) - - [6. MULTILINGUAL MODEL INFERENCE](MULTILINGUAL_MODEL_INFERENCE) + - [4. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY](#USING_CUSTOM_CHARACTERS) + - [5. MULTILINGUAL MODEL INFERENCE](MULTILINGUAL_MODEL_INFERENCE) - [ANGLE CLASSIFICATION MODEL INFERENCE](#ANGLE_CLASS_MODEL_INFERENCE) - [1. ANGLE CLASSIFICATION MODEL INFERENCE](#ANGLE_CLASS_MODEL_INFERENCE) @@ -44,26 +43,27 @@ Next, we first introduce how to convert a trained model into an inference model, Download the lightweight Chinese detection model: ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v1.1_det_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_det_train.tar -C ./ch_lite/ ``` The above model is a DB algorithm trained with MobileNetV3 as the backbone. To convert the trained model into an inference model, just run the following command: ``` # -c Set the training algorithm yml configuration file # -o Set optional parameters -# Global.checkpoints parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# Global.save_inference_dir Set the address where the converted model will be saved. +# Global.pretrained_model parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. +# Global.load_static_weights needs to be set to False +# Global.save_inference_dir Set the address where the converted model will be saved. -python3 tools/export_model.py -c configs/det/det_mv3_db_v1.1.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_det_train/best_accuracy Global.save_inference_dir=./inference/det_db/ +python3 tools/export_model.py -c configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_det_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_db/ ``` -When converting to an inference model, the configuration file used is the same as the configuration file used during training. In addition, you also need to set the `Global.checkpoints` and `Global.save_inference_dir` parameters in the configuration file. -`Global.checkpoints` points to the model parameter file saved during training, and `Global.save_inference_dir` is the directory where the generated inference model is saved. -After the conversion is successful, there are two files in the `save_inference_dir` directory: +When converting to an inference model, the configuration file used is the same as the configuration file used during training. In addition, you also need to set the `Global.pretrained_model` parameter in the configuration file. +After the conversion is successful, there are three files in the model save directory: ``` inference/det_db/ - └─ model Check the program file of inference model - └─ params Check the parameter file of the inference model + ├── inference.pdiparams # The parameter file of detection inference model + ├── inference.pdiparams.info # The parameter information of detection inference model, which can be ignored + └── inference.pdmodel # The program file of detection inference model ``` @@ -71,26 +71,28 @@ inference/det_db/ Download the lightweight Chinese recognition model: ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_train.tar && tar xf ch_ppocr_mobile_v1.1_rec_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_rec_train.tar -C ./ch_lite/ ``` The recognition model is converted to the inference model in the same way as the detection, as follows: ``` # -c Set the training algorithm yml configuration file # -o Set optional parameters -# Global.checkpoints parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# Global.save_inference_dir Set the address where the converted model will be saved. +# Global.pretrained_model parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. +# Global.load_static_weights needs to be set to False +# Global.save_inference_dir Set the address where the converted model will be saved. -python3 tools/export_model.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_rec_train/best_accuracy \ +python3 tools/export_model.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_rec_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/rec_crnn/ ``` If you have a model trained on your own dataset with a different dictionary file, please make sure that you modify the `character_dict_path` in the configuration file to your dictionary file path. -After the conversion is successful, there are two files in the directory: +After the conversion is successful, there are three files in the model save directory: ``` -/inference/rec_crnn/ - └─ model Identify the saved model files - └─ params Identify the parameter files of the inference model +inference/det_db/ + ├── inference.pdiparams # The parameter file of recognition inference model + ├── inference.pdiparams.info # The parameter information of recognition inference model, which can be ignored + └── inference.pdmodel # The program file of recognition model ``` @@ -98,25 +100,26 @@ After the conversion is successful, there are two files in the directory: Download the angle classification model: ``` -wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v1.1_cls_train.tar -C ./ch_lite/ +wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_cls_train.tar -C ./ch_lite/ ``` The angle classification model is converted to the inference model in the same way as the detection, as follows: ``` # -c Set the training algorithm yml configuration file # -o Set optional parameters -# Global.checkpoints parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# Global.save_inference_dir Set the address where the converted model will be saved. +# Global.pretrained_model parameter Set the training model address to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. +# Global.load_static_weights needs to be set to False +# Global.save_inference_dir Set the address where the converted model will be saved. -python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.checkpoints=./ch_lite/ch_ppocr_mobile_v1.1_cls_train/best_accuracy \ - Global.save_inference_dir=./inference/cls/ +python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_cls_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/cls/ ``` After the conversion is successful, there are two files in the directory: ``` -/inference/cls/ - └─ model Identify the saved model files - └─ params Identify the parameter files of the inference model +inference/det_db/ + ├── inference.pdiparams # The parameter file of angle class inference model + ├── inference.pdiparams.info # The parameter information of angle class inference model, which can be ignored + └── inference.pdmodel # The program file of angle class model ``` @@ -139,10 +142,12 @@ The visual text detection results are saved to the ./inference_results folder by ![](../imgs_results/det_res_2.jpg) -By setting the size of the parameter `det_max_side_len`, the maximum value of picture normalization in the detection algorithm is changed. When the length and width of the picture are less than det_max_side_len, the original picture is used for prediction, otherwise the picture is scaled to the maximum value for prediction. This parameter is set to det_max_side_len=960 by default. If the resolution of the input picture is relatively large and you want to use a larger resolution for prediction, you can execute the following command: +The size of the image is limited by the parameters `limit_type` and `det_limit_side_len`, `limit_type=max` is to limit the length of the long side <`det_limit_side_len`, and `limit_type=min` is to limit the length of the short side>`det_limit_side_len`, +When the picture does not meet the restriction conditions (for `limit_type=max`and long side >`det_limit_side_len` or for `min` and short side <`det_limit_side_len`), the image will be scaled proportionally. +This parameter is set to `limit_type='max', det_max_side_len=960` by default. If the resolution of the input picture is relatively large, and you want to use a larger resolution prediction, you can execute the following command: ``` -python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_max_side_len=1200 +python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_dir="./inference/det_db/" --det_limit_type=max --det_limit_side_len=1200 ``` If you want to use the CPU for prediction, execute the command as follows @@ -153,14 +158,10 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs/2.jpg" --det_model_di ### 2. DB TEXT DETECTION MODEL INFERENCE -First, convert the model saved in the DB text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link](https://paddleocr.bj.bcebos.com/det_r50_vd_db.tar)), you can use the following command to convert: +First, convert the model saved in the DB text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_r50_vd_db_v2.0_train.tar)), you can use the following command to convert: ``` -# Set the yml configuration file of the training algorithm after -c -# The Global.checkpoints parameter sets the address of the training model to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# The Global.save_inference_dir parameter sets the address where the converted model will be saved. - -python3 tools/export_model.py -c configs/det/det_r50_vd_db.yml -o Global.checkpoints="./models/det_r50_vd_db/best_accuracy" Global.save_inference_dir="./inference/det_db" +python3 tools/export_model.py -c configs/det/det_r50_vd_db.yml -o Global.pretrained_model=./det_r50_vd_db_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_db ``` DB text detection model inference, you can execute the following command: @@ -178,16 +179,11 @@ The visualized text detection results are saved to the `./inference_results` fol ### 3. EAST TEXT DETECTION MODEL INFERENCE -First, convert the model saved in the EAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link](https://paddleocr.bj.bcebos.com/det_r50_vd_east.tar)), you can use the following command to convert: +First, convert the model saved in the EAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link (coming soon)](link)), you can use the following command to convert: ``` -# Set the yml configuration file of the training algorithm after -c -# The Global.checkpoints parameter sets the address of the training model to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# The Global.save_inference_dir parameter sets the address where the converted model will be saved. - -python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.checkpoints="./models/det_r50_vd_east/best_accuracy" Global.save_inference_dir="./inference/det_east" +python3 tools/export_model.py -c configs/det/det_r50_vd_east.yml -o Global.pretrained_model=./det_r50_vd_east_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_east ``` - **For EAST text detection model inference, you need to set the parameter ``--det_algorithm="EAST"``**, run the following command: ``` @@ -196,7 +192,7 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs_en/img_10.jpg" --det_ The visualized text detection results are saved to the `./inference_results` folder by default, and the name of the result file is prefixed with 'det_res'. Examples of results are as follows: -![](../imgs_results/det_res_img_10_east.jpg) +(coming soon) **Note**: EAST post-processing locality aware NMS has two versions: Python and C++. The speed of C++ version is obviously faster than that of Python version. Due to the compilation version problem of NMS of C++ version, C++ version NMS will be called only in Python 3.5 environment, and python version NMS will be called in other cases. @@ -204,10 +200,10 @@ The visualized text detection results are saved to the `./inference_results` fol ### 4. SAST TEXT DETECTION MODEL INFERENCE #### (1). Quadrangle text detection model (ICDAR2015) -First, convert the model saved in the SAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_icdar2015.tar)), you can use the following command to convert: +First, convert the model saved in the SAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the ICDAR2015 English dataset as an example ([model download link (coming soon)](link)), you can use the following command to convert: ``` -python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.checkpoints="./models/sast_r50_vd_icdar2015/best_accuracy" Global.save_inference_dir="./inference/det_sast_ic15" +python3 tools/export_model.py -c configs/det/det_r50_vd_sast_icdar15.yml -o Global.pretrained_model=./det_r50_vd_sast_icdar15_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_ic15 ``` **For SAST quadrangle text detection model inference, you need to set the parameter `--det_algorithm="SAST"`**, run the following command: @@ -218,13 +214,13 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img The visualized text detection results are saved to the `./inference_results` folder by default, and the name of the result file is prefixed with 'det_res'. Examples of results are as follows: -![](../imgs_results/det_res_img_10_sast.jpg) +(coming soon) #### (2). Curved text detection model (Total-Text) -First, convert the model saved in the SAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the Total-Text English dataset as an example ([model download link](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar)), you can use the following command to convert: +First, convert the model saved in the SAST text detection training process into an inference model. Taking the model based on the Resnet50_vd backbone network and trained on the Total-Text English dataset as an example ([model download link (coming soon)](https://paddleocr.bj.bcebos.com/SAST/sast_r50_vd_total_text.tar)), you can use the following command to convert: ``` -python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.checkpoints="./models/sast_r50_vd_total_text/best_accuracy" Global.save_inference_dir="./inference/det_sast_tt" +python3 tools/export_model.py -c configs/det/det_r50_vd_sast_totaltext.yml -o Global.pretrained_model=./det_r50_vd_sast_totaltext_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/det_sast_tt ``` **For SAST curved text detection model inference, you need to set the parameter `--det_algorithm="SAST"` and `--det_sast_polygon=True`**, run the following command: @@ -235,7 +231,7 @@ python3 tools/infer/predict_det.py --det_algorithm="SAST" --image_dir="./doc/img The visualized text detection results are saved to the `./inference_results` folder by default, and the name of the result file is prefixed with 'det_res'. Examples of results are as follows: -![](../imgs_results/det_res_img623_sast.jpg) +(coming soon) **Note**: SAST post-processing locality aware NMS has two versions: Python and C++. The speed of C++ version is obviously faster than that of Python version. Due to the compilation version problem of NMS of C++ version, C++ version NMS will be called only in Python 3.5 environment, and python version NMS will be called in other cases. @@ -258,25 +254,22 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/ch/word_4.jpg" After executing the command, the prediction results (recognized text and score) of the above image will be printed on the screen. -Predicts of ./doc/imgs_words/ch/word_4.jpg:['实力活力', 0.89552695] - +```bash +Predicts of ./doc/imgs_words/ch/word_4.jpg:('实力活力', 0.98458153) +``` ### 2. CTC-BASED TEXT RECOGNITION MODEL INFERENCE -Taking STAR-Net as an example, we introduce the recognition model inference based on CTC loss. CRNN and Rosetta are used in a similar way, by setting the recognition algorithm parameter `rec_algorithm`. +Taking CRNN as an example, we introduce the recognition model inference based on CTC loss. Rosetta and Star-Net are used in a similar way, No need to set the recognition algorithm parameter rec_algorithm. -First, convert the model saved in the STAR-Net text recognition training process into an inference model. Taking the model based on Resnet34_vd backbone network, using MJSynth and SynthText (two English text recognition synthetic datasets) for training, as an example ([model download address](https://paddleocr.bj.bcebos.com/rec_r34_vd_tps_bilstm_ctc.tar)). It can be converted as follow: +First, convert the model saved in the CRNN text recognition training process into an inference model. Taking the model based on Resnet34_vd backbone network, using MJSynth and SynthText (two English text recognition synthetic datasets) for training, as an example ([model download address](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_none_bilstm_ctc_v2.0_train.tar)). It can be converted as follow: ``` -# Set the yml configuration file of the training algorithm after -c -# The Global.checkpoints parameter sets the address of the training model to be converted without adding the file suffix .pdmodel, .pdopt or .pdparams. -# The Global.save_inference_dir parameter sets the address where the converted model will be saved. - -python3 tools/export_model.py -c configs/rec/rec_r34_vd_tps_bilstm_ctc.yml -o Global.checkpoints="./models/rec_r34_vd_tps_bilstm_ctc/best_accuracy" Global.save_inference_dir="./inference/starnet" +python3 tools/export_model.py -c configs/det/rec_r34_vd_none_bilstm_ctc.yml -o Global.pretrained_model=./rec_r34_vd_none_bilstm_ctc_v2.0_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/rec_crnn ``` -For STAR-Net text recognition model inference, execute the following commands: +For CRNN text recognition model inference, execute the following commands: ``` python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/starnet/" --rec_image_shape="3, 32, 100" --rec_char_type="en" @@ -284,12 +277,20 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png ### 3. ATTENTION-BASED TEXT RECOGNITION MODEL INFERENCE -![](../imgs_words_en/word_336.png) +The recognition model based on Attention loss is different from ctc, and additional recognition algorithm parameters need to be set --rec_algorithm="RARE" After executing the command, the recognition result of the above image is as follows: +```bash +python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./inference/rare/" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_algorithm="RARE" +``` -Predicts of ./doc/imgs_words_en/word_336.png:['super', 0.9999555] +![](../imgs_words_en/word_336.png) +After executing the command, the recognition result of the above image is as follows: + +```bash +Predicts of ./doc/imgs_words_en/word_336.png:('super', 0.9999073) +``` **Note**：Since the above model refers to [DTRB](https://arxiv.org/abs/1904.01906) text recognition training and evaluation process, it is different from the training of lightweight Chinese recognition model in two aspects: - The image resolution used in training is different: the image resolution used in training the above model is [3，32，100], while during our Chinese model training, in order to ensure the recognition effect of long text, the image resolution used in training is [3, 32, 320]. The default shape parameter of the inference stage is the image resolution used in training phase, that is [3, 32, 320]. Therefore, when running inference of the above English model here, you need to set the shape of the recognition image through the parameter `rec_image_shape`. @@ -301,31 +302,16 @@ self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz" dict_character = list(self.character_str) ``` - -### 4. SRN-BASED TEXT RECOGNITION MODEL INFERENCE - -The recognition model based on SRN requires additional setting of the recognition algorithm parameter --rec_algorithm="SRN". -At the same time, it is necessary to ensure that the predicted shape is consistent with the training, such as: --rec_image_shape="1, 64, 256" - -``` -python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" \ - --rec_model_dir="./inference/srn/" \ - --rec_image_shape="1, 64, 256" \ - --rec_char_type="en" \ - --rec_algorithm="SRN" -``` - - -### 5. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY -If the chars dictionary is modified during training, you need to specify the new dictionary path by setting the parameter `rec_char_dict_path` when using your inference model to predict. +### 4. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY +If the text dictionary is modified during training, when using the inference model to predict, you need to specify the dictionary path used by `--rec_char_dict_path`, and set `rec_char_type=ch` ``` -python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="en" --rec_char_dict_path="your text dict path" +python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="ch" --rec_char_dict_path="your text dict path" ``` -### 6. MULTILINGAUL MODEL INFERENCE +### 5. MULTILINGAUL MODEL INFERENCE If you need to predict other language models, when using inference model prediction, you need to specify the dictionary path used by `--rec_char_dict_path`. At the same time, in order to get the correct visualization results, You need to specify the visual font path through `--vis_font_path`. There are small language fonts provided by default under the `doc/` path, such as Korean recognition: @@ -337,9 +323,7 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/korean/1.jpg" - After executing the command, the prediction result of the above figure is: ``` text -2020-09-19 16:15:05,076-INFO: index: [205 206 38 39] -2020-09-19 16:15:05,077-INFO: word : 바탕으로 -2020-09-19 16:15:05,077-INFO: score: 0.9171358942985535 +Predicts of ./doc/imgs_words/korean/1.jpg:('바탕으로', 0.9948904) ``` @@ -354,15 +338,16 @@ The following will introduce the angle classification model inference. For angle classification model inference, you can execute the following commands: ``` -python3 tools/infer/predict_cls.py --image_dir="./doc/imgs_words/ch/word_4.jpg" --cls_model_dir="./inference/cls/" +python3 tools/infer/predict_cls.py --image_dir="./doc/imgs_words_en/word_10.png" --cls_model_dir="./inference/cls/" ``` -![](../imgs_words/ch/word_4.jpg) +![](../imgs_words_en/word_10.png) After executing the command, the prediction results (classification angle and score) of the above image will be printed on the screen. -Predicts of ./doc/imgs_words/ch/word_4.jpg:['0', 0.9999963] - +``` + Predicts of ./doc/imgs_words_en/word_10.png:['0', 0.9999995] +``` ## TEXT DETECTION ANGLE CLASSIFICATION AND RECOGNITION INFERENCE CONCATENATION @@ -399,4 +384,4 @@ python3 tools/infer/predict_system.py --image_dir="./doc/imgs_en/img_10.jpg" --d After executing the command, the recognition result image is as follows: -![](../imgs_results/img_10.jpg) +(coming soon) diff --git a/doc/doc_en/models_list_en.md b/doc/doc_en/models_list_en.md index 3d3cdc1d6c2e2a52c64f6ed0fd502456271173b6..63d8c598bbe4e3b37ae47804e595438ee79905c8 100644 --- a/doc/doc_en/models_list_en.md +++ b/doc/doc_en/models_list_en.md @@ -1,4 +1,4 @@ -## OCR model list（V1.1, updated on 9.22） +## OCR model list（V1.1, updated on 2020.12.12） - [1. Text Detection Model](#Detection) - [2. Text Recognition Model](#Recognition) @@ -10,61 +10,62 @@ The downloadable models provided by PaddleOCR include `inference model`, `trained model`, `pre-trained model` and `slim model`. The differences between the models are as follows: |model type|model format|description| -|-|-|-| -|inference model|model、params|Used for reasoning based on Python prediction engine. [detail](./inference_en.md)| -|trained model / pre-trained model|\*.pdmodel、\*.pdopt、\*.pdparams|The checkpoints model saved in the training process, which stores the parameters of the model, mostly used for model evaluation and continuous training.| +|--- | --- | --- | +|inference model|inference.pdmodel、inference.pdiparams|Used for reasoning based on Python prediction engine，[detail](./inference_en.md)| +|trained model, pre-trained model|\*.pdparams、\*.pdopt、\*.states |The checkpoints model saved in the training process, which stores the parameters of the model, mostly used for model evaluation and continuous training.| |slim model|\*.nb|Generally used for Lite deployment| - ### 1. Text Detection Model -|model name|description|config|model size|download| -|-|-|-|-|-| -|ch_ppocr_mobile_slim_v1.1_det|Slim pruned lightweight model, supporting Chinese, English, multilingual text detection|[det_mv3_db_v1.1.yml](../../configs/det/det_mv3_db_v1.1.yml)|1.4M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/det/ch_ppocr_mobile_v1.1_det_prune_infer.tar) / [slim model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_det_prune_opt.nb)| -|ch_ppocr_mobile_v1.1_det|Original lightweight model, supporting Chinese, English, multilingual text detection|[det_mv3_db_v1.1.yml](../../configs/det/det_mv3_db_v1.1.yml)|2.6M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar)| -|ch_ppocr_server_v1.1_det|General model, which is larger than the lightweight model, but achieved better performance|[det_r18_vd_db_v1.1.yml](../../configs/det/det_r18_vd_db_v1.1.yml)|47.2M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_train.tar)| +|model name|description|config|model size|download| +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_det|Slim pruned lightweight model, supporting Chinese, English, multilingual text detection|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)| |[inference model (coming soon)](link) / [slim model (coming soon)](link)| +|ch_ppocr_mobile_v2.0_det|Original lightweight model, supporting Chinese, English, multilingual text detection|[ch_det_mv3_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)|3M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)| +|ch_ppocr_server_v2.0_det|General model, which is larger than the lightweight model, but achieved better performance|[ch_det_res18_db_v2.0.yml](../../configs/det/ch_ppocr_v2.0/ch_det_res18_db_v2.0.yml)|47M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar)| ### 2. Text Recognition Model #### Chinese Recognition Model + |model name|description|config|model size|download| -|-|-|-|-|-| -|ch_ppocr_mobile_slim_v1.1_rec|Slim pruned and quantized lightweight model, supporting Chinese, English and number recognition|[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml)|1.6M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/rec/ch_ppocr_mobile_v1.1_rec_quant_infer.tar) / [slim model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_rec_quant_opt.nb) | -|ch_ppocr_mobile_v1.1_rec|Original lightweight model, supporting Chinese, English and number recognition|[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml)|4.6M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_train.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_pre.tar) | -|ch_ppocr_server_v1.1_rec|General model, supporting Chinese, English and number recognition|[rec_chinese_common_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yml)|105M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_train.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_pre.tar) | +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_rec|Slim pruned and quantized lightweight model, supporting Chinese, English and number recognition|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)| |[inference model (coming soon)](link) / [slim model (coming soon)](link) | +|ch_ppocr_mobile_v2.0_rec|Original lightweight model, supporting Chinese, English and number recognition|[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)|3.71M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_train.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +|ch_ppocr_server_v2.0_rec|General model, supporting Chinese, English and number recognition|[rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml)|94.8M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_train.tar) / [pre-trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | + **Note:** The `trained model` is finetuned on the `pre-trained model` with real data and synthsized vertical text data, which achieved better performance in real scene. The `pre-trained model` is directly trained on the full amount of real data and synthsized data, which is more suitable for finetune on your own dataset. #### English Recognition Model + |model name|description|config|model size|download| -|-|-|-|-|-| -|en_ppocr_mobile_slim_v1.1_rec|Slim pruned and quantized lightweight model, supporting English and number recognition|[rec_en_lite_train.yml](../../configs/rec/multi_languages/rec_en_lite_train.yml)|0.9M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/en/en_ppocr_mobile_v1.1_rec_quant_infer.tar) / [slim model](https://paddleocr.bj.bcebos.com/20-09-22/mobile-slim/en/en_ppocr_mobile_v1.1_rec_quant_opt.nb) | -|en_ppocr_mobile_v1.1_rec|Original lightweight model, supporting English and number recognition|[rec_en_lite_train.yml](../../configs/rec/multi_languages/rec_en_lite_train.yml)|2.0M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/en/en_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/en/en_ppocr_mobile_v1.1_rec_train.tar) | +| --- | --- | --- | --- | --- | +|en_number_mobile_slim_v2.0_rec|Slim pruned and quantized lightweight model, supporting English and number recognition|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)| |[inference model (coming soon )](link) / [slim model (coming soon)](link) | +|en_number_mobile_v2.0_rec|Original lightweight model, supporting English and number recognition|[rec_en_number_lite_train.yml](../../configs/rec/multi_language/rec_en_number_lite_train.yml)|2.56M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_train.tar) | #### Multilingual Recognition Model（Updating...） -|model name|description|config|model size|download| -|-|-|-|-|-| -| french_ppocr_mobile_v1.1_rec |Lightweight model for French recognition|[rec_french_lite_train.yml](../../configs/rec/multi_languages/rec_french_lite_train.yml)|2.1M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_train.tar) | -| german_ppocr_mobile_v1.1_rec |German model for French recognition|[rec_ger_lite_train.yml](../../configs/rec/multi_languages/rec_ger_lite_train.yml)|2.1M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_train.tar) | -| korean_ppocr_mobile_v1.1_rec |Lightweight model for Korean recognition|[rec_korean_lite_train.yml](../../configs/rec/multi_languages/rec_korean_lite_train.yml)|3.4M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_train.tar) | -| japan_ppocr_mobile_v1.1_rec |Lightweight model for Japanese recognition|[rec_japan_lite_train.yml](../../configs/rec/multi_languages/rec_japan_lite_train.yml)|3.7M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_train.tar) | +|model name|description|config|model size|download| +| --- | --- | --- | --- | --- | +| french_mobile_v2.0_rec |Lightweight model for French recognition|[rec_french_lite_train.yml](../../configs/rec/multi_language/rec_french_lite_train.yml)|2.65M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_train.tar) | +| german_mobile_v2.0_rec |Lightweight model for French recognition|[rec_german_lite_train.yml](../../configs/rec/multi_language/rec_german_lite_train.yml)|2.65M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_train.tar) | +| korean_mobile_v2.0_rec |Lightweight model for Korean recognition|[rec_korean_lite_train.yml](../../configs/rec/multi_language/rec_korean_lite_train.yml)|3.9M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_train.tar) | +| japan_mobile_v2.0_rec |Lightweight model for Japanese recognition|[rec_japan_lite_train.yml](../../configs/rec/multi_language/rec_japan_lite_train.yml)|4.23M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_train.tar) | ### 3. Text Angle Classification Model + |model name|description|config|model size|download| -|-|-|-|-|-| -|ch_ppocr_mobile_v1.1_cls_quant|Slim quantized model|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|0.5M|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_quant_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_quant_train.tar) / [slim model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/lite/ch_ppocr_mobile_v1.1_cls_quant_opt.nb) | -|ch_ppocr_mobile_v1.1_cls|Original model|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|850kb|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) | +| --- | --- | --- | --- | --- | +|ch_ppocr_mobile_slim_v2.0_cls|Slim quantized model|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)| |[inference model (coming soon)](link) / [trained model](link) / [slim model](link) | +|ch_ppocr_mobile_v2.0_cls|Original model|[cls_mv3.yml](../../configs/cls/cls_mv3.yml)|1.38M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) | + +## OCR model list (V1.1，updated on 2020.9.22) -## OCR model list（V1.0, updated on 7.16） -|model name|description|detection model|recognition model|recognition model supporting space recognition| -|-|-|-|-|-| -|chinese_db_crnn_mobile|8.6M lightweight OCR model|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_mv3_db.tar) | [inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn.tar) |[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_mv3_crnn_enhance.tar) -|chinese_db_crnn_server|General OCR model|[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_det_r50_vd_db.tar) | [inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn.tar) |[inference model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/ch_models/ch_rec_r34_vd_crnn_enhance.tar) +[1.1 series model address](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/models_list.md) diff --git a/doc/doc_en/quickstart_en.md b/doc/doc_en/quickstart_en.md index 6a5f369349f3142c29dbe7edb766a51a0789cf37..e351ecc650d621b1da5f34dd941eaf6fb3094402 100644 --- a/doc/doc_en/quickstart_en.md +++ b/doc/doc_en/quickstart_en.md @@ -5,17 +5,17 @@ Please refer to [quick installation](./installation_en.md) to configure the PaddleOCR operating environment. -* Note: Support the use of PaddleOCR through whl package installation，pelease refer [PaddleOCR Package](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_en/whl_en.md). +* Note: Support the use of PaddleOCR through whl package installation，pelease refer [PaddleOCR Package](./whl_en.md). ## 2.inference models -The detection and recognition models on the mobile and server sides are as follows. For more models (including multiple languages), please refer to [PP-OCR v1.1 series model list](../doc_ch/models_list.md) +The detection and recognition models on the mobile and server sides are as follows. For more models (including multiple languages), please refer to [PP-OCR v2.0 series model list](../doc_ch/models_list.md) - -| Model introduction | Model name | Recommended scene | Detection model | Direction Classifier | Recognition model | +| Model introduction | Model name | Recommended scene | Detection model | Direction Classifier | Recognition model | | ------------ | --------------- | ----------------|---- | ---------- | -------- | -| Ultra-lightweight Chinese OCR model（8.1M） | ch_ppocr_mobile_v1.1_xx |Mobile-side/Server-side|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_train.tar)|[inference model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_pre.tar) | -| Universal Chinese OCR model（155.1M） |ch_ppocr_server_v1.1_xx|Server-side |[inference model](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/server/det/ch_ppocr_server_v1.1_det_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/20-09-22/server/rec/ch_ppocr_server_v1.1_rec_pre.tar) | +| Ultra-lightweight Chinese OCR model (8.1M) | ch_ppocr_mobile_v2.0_xx |Mobile-side/Server-side|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar)|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_pre.tar) | +| Universal Chinese OCR model (143M) | ch_ppocr_server_v2.0_xx |Server-side |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar) |[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar) / [pretrained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_pre.tar) | + * If `wget` is not installed in the windows environment, you can copy the link to the browser to download when downloading the model, then uncompress it and place it in the corresponding directory. @@ -37,46 +37,47 @@ Take the ultra-lightweight model as an example: ``` mkdir inference && cd inference # Download the detection model of the ultra-lightweight Chinese OCR model and uncompress it -wget https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar && tar xf ch_ppocr_mobile_v1.1_det_infer.tar +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_ppocr_mobile_v2.0_det_infer.tar # Download the recognition model of the ultra-lightweight Chinese OCR model and uncompress it -wget https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar && tar xf ch_ppocr_mobile_v1.1_rec_infer.tar -# Download the direction classifier model of the ultra-lightweight Chinese OCR model and uncompress it -wget https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar && tar xf ch_ppocr_mobile_v1.1_cls_infer.tar +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar +# Download the angle classifier model of the ultra-lightweight Chinese OCR model and uncompress it +wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar && tar xf ch_ppocr_mobile_v2.0_cls_infer.tar cd .. ``` After decompression, the file structure should be as follows: ``` -|-inference - |-ch_ppocr_mobile_v1.1_det_infer - |- model - |- params - |-ch_ppocr_mobile_v1.1_rec_infer - |- model - |- params - |-ch_ppocr_mobile_v1.1_cls_infer - |- model - |- params - ... +├── ch_ppocr_mobile_v2.0_cls_infer +│ ├── inference.pdiparams +│ ├── inference.pdiparams.info +│ └── inference.pdmodel +├── ch_ppocr_mobile_v2.0_det_infer +│ ├── inference.pdiparams +│ ├── inference.pdiparams.info +│ └── inference.pdmodel +├── ch_ppocr_mobile_v2.0_rec_infer + ├── inference.pdiparams + ├── inference.pdiparams.info + └── inference.pdmodel ``` ## 3. Single image or image set prediction -* The following code implements text detection and recognition process. When performing prediction, you need to specify the path of a single image or image set through the parameter `image_dir`, the parameter `det_model_dir` specifies the path to detect the inference model, the parameter `rec_model_dir` specifies the path to identify the inference model, the parameter `use_angle_cls` specifies whether to use the direction classifier, the parameter `cls_model_dir` specifies the path to identify the direction classifier model, the parameter `use_space_char` specifies whether to predict the space char. The visual results are saved to the `./inference_results` folder by default. +* The following code implements text detection、angle class and recognition process. When performing prediction, you need to specify the path of a single image or image set through the parameter `image_dir`, the parameter `det_model_dir` specifies the path to detect the inference model, the parameter `rec_model_dir` specifies the path to identify the inference model, the parameter `use_angle_cls` specifies whether to use the direction classifier, the parameter `cls_model_dir` specifies the path to identify the direction classifier model, the parameter `use_space_char` specifies whether to predict the space char. The visual results are saved to the `./inference_results` folder by default. ```bash # Predict a single image specified by image_dir -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True # Predict imageset specified by image_dir -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True # If you want to use the CPU for prediction, you need to set the use_gpu parameter to False -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True --use_gpu=False +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_mobile_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_mobile_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True --use_gpu=False ``` - Universal Chinese OCR model @@ -85,7 +86,7 @@ Please follow the above steps to download the corresponding models and update th ``` # Predict a single image specified by image_dir -python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_server_v1.1_det_infer/" --rec_model_dir="./inference/ch_ppocr_server_v1.1_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v1.1_cls_infer/" --use_angle_cls=True --use_space_char=True +python3 tools/infer/predict_system.py --image_dir="./doc/imgs/11.jpg" --det_model_dir="./inference/ch_ppocr_server_v2.0_det_infer/" --rec_model_dir="./inference/ch_ppocr_server_v2.0_rec_infer/" --cls_model_dir="./inference/ch_ppocr_mobile_v2.0_cls_infer/" --use_angle_cls=True --use_space_char=True ``` * Note diff --git a/doc/doc_en/recognition_en.md b/doc/doc_en/recognition_en.md index 41b00c52a7780d02c144c251553f427e5b875e5e..1539b288da2518bf5441adea7983135f3c46619f 100644 --- a/doc/doc_en/recognition_en.md +++ b/doc/doc_en/recognition_en.md @@ -114,11 +114,14 @@ In `word_dict.txt`, there is a single word in each line, which maps characters a `ppocr/utils/dict/french_dict.txt` is a French dictionary with 118 characters -`ppocr/utils/dict/japan_dict.txt` is a French dictionary with 4399 characters +`ppocr/utils/dict/japan_dict.txt` is a Japanese dictionary with 4399 characters -`ppocr/utils/dict/korean_dict.txt` is a French dictionary with 3636 characters +`ppocr/utils/dict/korean_dict.txt` is a Korean dictionary with 3636 characters + +`ppocr/utils/dict/german_dict.txt` is a German dictionary with 131 characters + +`ppocr/utils/dict/en_dict.txt` is a English dictionary with 63 characters -`ppocr/utils/dict/german_dict.txt` is a French dictionary with 131 characters You can use it on demand. @@ -135,7 +138,7 @@ If you need to customize dic file, please add character_dict_path field in confi - Add space category -If you want to support the recognition of the `space` category, please set the `use_space_char` field in the yml file to `true`. +If you want to support the recognition of the `space` category, please set the `use_space_char` field in the yml file to `True`. **Note: use_space_char only takes effect when character_type=ch** @@ -149,19 +152,18 @@ First download the pretrain model, you can download the trained model to finetun ``` cd PaddleOCR/ # Download the pre-trained model of MobileNetV3 -wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/rec_mv3_none_bilstm_ctc.tar +wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_bilstm_ctc_v2.0_train.tar # Decompress model parameters cd pretrain_models -tar -xf rec_mv3_none_bilstm_ctc.tar && rm -rf rec_mv3_none_bilstm_ctc.tar +tar -xf rec_mv3_none_bilstm_ctc_v2.0_train.tar && rm -rf rec_mv3_none_bilstm_ctc_v2.0_train.tar ``` Start training: ``` -# GPU training Support single card and multi-card training, specify the card number through CUDA_VISIBLE_DEVICES -export CUDA_VISIBLE_DEVICES=0,1,2,3 +# GPU training Support single card and multi-card training, specify the card number through --gpus # Training icdar15 English data and saving the log as train_rec.log -python3 tools/train.py -c configs/rec/rec_icdar15_train.yml 2>&1 | tee train_rec.log +python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/rec/rec_icdar15_train.yml ``` - Data Augmentation @@ -184,8 +186,8 @@ If the evaluation set is large, the test will be time-consuming. It is recommend | Configuration file | Algorithm | backbone | trans | seq | pred | | :--------: | :-------: | :-------: | :-------: | :-----: | :-----: | -| [rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml) | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | -| [rec_chinese_common_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_common_train_v1.1.yml) | CRNN | ResNet34_vd | None | BiLSTM | ctc | +| [rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml) | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | +| [rec_chinese_common_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_common_train_v2.0.yml) | CRNN | ResNet34_vd | None | BiLSTM | ctc | | rec_chinese_lite_train.yml | CRNN | Mobilenet_v3 small 0.5 | None | BiLSTM | ctc | | rec_chinese_common_train.yml | CRNN | ResNet34_vd | None | BiLSTM | ctc | | rec_icdar15_train.yml | CRNN | Mobilenet_v3 large 0.5 | None | BiLSTM | ctc | @@ -195,43 +197,72 @@ If the evaluation set is large, the test will be time-consuming. It is recommend | rec_mv3_tps_bilstm_attn.yml | RARE | Mobilenet_v3 large 0.5 | tps | BiLSTM | attention | | rec_r34_vd_none_bilstm_ctc.yml | CRNN | Resnet34_vd | None | BiLSTM | ctc | | rec_r34_vd_none_none_ctc.yml | Rosetta | Resnet34_vd | None | None | ctc | -| rec_r34_vd_tps_bilstm_attn.yml | RARE | Resnet34_vd | tps | BiLSTM | attention | | rec_r34_vd_tps_bilstm_ctc.yml | STARNet | Resnet34_vd | tps | BiLSTM | ctc | For training Chinese data, it is recommended to use -训练中文数据，推荐使用[rec_chinese_lite_train_v1.1.yml](../../configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml). If you want to try the result of other algorithms on the Chinese data set, please refer to the following instructions to modify the configuration file: +[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml). If you want to try the result of other algorithms on the Chinese data set, please refer to the following instructions to modify the configuration file: co -Take `rec_mv3_none_none_ctc.yml` as an example: +Take `rec_chinese_lite_train_v2.0.yml` as an example: ``` Global: ... - # Modify image_shape to fit long text - image_shape: [3, 32, 320] - ... + # Add a custom dictionary, such as modify the dictionary, please point the path to the new dictionary + character_dict_path: ppocr/utils/ppocr_keys_v1.txt # Modify character type character_type: ch - # Add a custom dictionary, such as modify the dictionary, please point the path to the new dictionary - character_dict_path: ./ppocr/utils/ppocr_keys_v1.txt ... - # Modify reader type - reader_yml: ./configs/rec/rec_chinese_reader.yml - # Whether to use data augmentation - distort: true # Whether to recognize spaces - use_space_char: true - ... + use_space_char: True -... Optimizer: ... # Add learning rate decay strategy - decay: - function: cosine_decay - # Each epoch contains iter number - step_each_epoch: 20 - # Total epoch number - total_epoch: 1000 + lr: + name: Cosine + learning_rate: 0.001 + ... + +... + +Train: + dataset: + # Type of dataset，we support LMDBDateSet and SimpleDataSet + name: SimpleDataSet + # Path of dataset + data_dir: ./train_data/ + # Path of train list + label_file_list: ["./train_data/train_list.txt"] + transforms: + ... + - RecResizeImg: + # Modify image_shape to fit long text + image_shape: [3, 32, 320] + ... + loader: + ... + # Train batch_size for Single card + batch_size_per_card: 256 + ... + +Eval: + dataset: + # Type of dataset，we support LMDBDateSet and SimpleDataSet + name: SimpleDataSet + # Path of dataset + data_dir: ./train_data + # Path of eval list + label_file_list: ["./train_data/val_list.txt"] + transforms: + ... + - RecResizeImg: + # Modify image_shape to fit long text + image_shape: [3, 32, 320] + ... + loader: + # Eval batch_size for Single card + batch_size_per_card: 256 + ... ``` **Note that the configuration file for prediction/evaluation must be consistent with the training.** @@ -257,18 +288,33 @@ Take `rec_french_lite_train` as an example: ``` Global: ... - # Add a custom dictionary, if you modify the dictionary - # please point the path to the new dictionary + # Add a custom dictionary, such as modify the dictionary, please point the path to the new dictionary character_dict_path: ./ppocr/utils/dict/french_dict.txt - # Add data augmentation during training - distort: true - # Identify spaces - use_space_char: true - ... - # Modify reader type - reader_yml: ./configs/rec/multi_languages/rec_french_reader.yml ... + # Whether to recognize spaces + use_space_char: True + ... + +Train: + dataset: + # Type of dataset，we support LMDBDateSet and SimpleDataSet + name: SimpleDataSet + # Path of dataset + data_dir: ./train_data/ + # Path of train list + label_file_list: ["./train_data/french_train.txt"] + ... + +Eval: + dataset: + # Type of dataset，we support LMDBDateSet and SimpleDataSet + name: SimpleDataSet + # Path of dataset + data_dir: ./train_data + # Path of eval list + label_file_list: ["./train_data/french_val.txt"] + ... ``` @@ -277,9 +323,8 @@ Global: The evaluation data set can be modified via `configs/rec/rec_icdar15_reader.yml` setting of `label_file_path` in EvalReader. ``` -export CUDA_VISIBLE_DEVICES=0 # GPU evaluation, Global.checkpoints is the weight to be tested -python3 tools/eval.py -c configs/rec/rec_icdar15_reader.yml -o Global.checkpoints={path/to/weights}/best_accuracy +python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_icdar15_reader.yml -o Global.checkpoints={path/to/weights}/best_accuracy ``` @@ -294,7 +339,7 @@ The default prediction picture is stored in `infer_img`, and the weight is speci ``` # Predict English results -python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml -o Global.checkpoints={path/to/weights}/best_accuracy TestReader.infer_img=doc/imgs_words/en/word_1.jpg +python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.checkpoints={path/to/weights}/best_accuracy TestReader.infer_img=doc/imgs_words/en/word_1.jpg ``` Input image: @@ -309,11 +354,11 @@ infer_img: doc/imgs_words/en/word_1.png word : joint ``` -The configuration file used for prediction must be consistent with the training. For example, you completed the training of the Chinese model with `python3 tools/train.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml`, you can use the following command to predict the Chinese model: +The configuration file used for prediction must be consistent with the training. For example, you completed the training of the Chinese model with `python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml`, you can use the following command to predict the Chinese model: ``` # Predict Chinese results -python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v1.1/rec_chinese_lite_train_v1.1.yml -o Global.checkpoints={path/to/weights}/best_accuracy TestReader.infer_img=doc/imgs_words/ch/word_1.jpg +python3 tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml -o Global.checkpoints={path/to/weights}/best_accuracy TestReader.infer_img=doc/imgs_words/ch/word_1.jpg ``` Input image: diff --git a/doc/imgs_results/2.jpg b/doc/imgs_results/2.jpg index 201ef9ee492702118cd1638ed8a0b832c1c6d9ed..99f7e63b02556506dadf8d838eee22534d21d82c 100644 Binary files a/doc/imgs_results/2.jpg and b/doc/imgs_results/2.jpg differ diff --git a/doc/imgs_results/det_res_2.jpg b/doc/imgs_results/det_res_2.jpg index aebcd8ccaca02db7ed4a09cd63ade422abc4735f..c0ae501a7aff7807f53b743745005653775b0d03 100644 Binary files a/doc/imgs_results/det_res_2.jpg and b/doc/imgs_results/det_res_2.jpg differ diff --git a/doc/imgs_results/det_res_img_10_db.jpg b/doc/imgs_results/det_res_img_10_db.jpg index bde1585cb50137ae1fd33ce7edfa59e7224ddc96..6af89f6bb32191c361c439c9d26e0239b5392fd9 100644 Binary files a/doc/imgs_results/det_res_img_10_db.jpg and b/doc/imgs_results/det_res_img_10_db.jpg differ diff --git a/doc/joinus.PNG b/doc/joinus.PNG index fa11f286d7d2d56d18d94e9034c3be77c974d42f..a6e947489831d90a841c3bb6f21596d5dac7e1ac 100644 Binary files a/doc/joinus.PNG and b/doc/joinus.PNG differ diff --git a/paddleocr.py b/paddleocr.py index 17306e79fe3c94e7f885de408a6c0f6c060a67da..1d8cd254644af77ea965d3fb5905f87a9b141e52 100644 --- a/paddleocr.py +++ b/paddleocr.py @@ -35,44 +35,45 @@ __all__ = ['PaddleOCR'] model_urls = { 'det': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/det/ch_ppocr_mobile_v1.1_det_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar', 'rec': { 'ch': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/rec/ch_ppocr_mobile_v1.1_rec_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar', 'dict_path': './ppocr/utils/ppocr_keys_v1.txt' }, 'en': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/en/en_ppocr_mobile_v1.1_rec_infer.tar', - 'dict_path': './ppocr/utils/ic15_dict.txt' + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/en_number_mobile_v2.0_rec_infer.tar', + 'dict_path': './ppocr/utils/dict/en_dict.txt' }, 'french': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/fr/french_ppocr_mobile_v1.1_rec_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_infer.tar', 'dict_path': './ppocr/utils/dict/french_dict.txt' }, 'german': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/ge/german_ppocr_mobile_v1.1_rec_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_infer.tar', 'dict_path': './ppocr/utils/dict/german_dict.txt' }, 'korean': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/kr/korean_ppocr_mobile_v1.1_rec_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_infer.tar', 'dict_path': './ppocr/utils/dict/korean_dict.txt' }, 'japan': { 'url': - 'https://paddleocr.bj.bcebos.com/20-09-22/mobile/jp/japan_ppocr_mobile_v1.1_rec_infer.tar', + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_infer.tar', 'dict_path': './ppocr/utils/dict/japan_dict.txt' } }, 'cls': - 'https://paddleocr.bj.bcebos.com/20-09-22/cls/ch_ppocr_mobile_v1.1_cls_infer.tar' + 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar' } SUPPORT_DET_MODEL = ['DB'] +VERSION = 2.0 SUPPORT_REC_MODEL = ['CRNN'] BASE_DIR = os.path.expanduser("~/.paddleocr/") @@ -94,20 +95,24 @@ def download_with_progressbar(url, save_path): def maybe_download(model_storage_directory, url): # using custom model - if not os.path.exists(os.path.join( - model_storage_directory, 'model')) or not os.path.exists( - os.path.join(model_storage_directory, 'params')): + tar_file_name_list = [ + 'inference.pdiparams', 'inference.pdiparams.info', 'inference.pdmodel' + ] + if not os.path.exists( + os.path.join(model_storage_directory, 'inference.pdiparams') + ) or not os.path.exists( + os.path.join(model_storage_directory, 'inference.pdmodel')): tmp_path = os.path.join(model_storage_directory, url.split('/')[-1]) print('download {} to {}'.format(url, tmp_path)) os.makedirs(model_storage_directory, exist_ok=True) download_with_progressbar(url, tmp_path) with tarfile.open(tmp_path, 'r') as tarObj: for member in tarObj.getmembers(): - if "model" in member.name: - filename = 'model' - elif "params" in member.name: - filename = 'params' - else: + filename = None + for tar_file_name in tar_file_name_list: + if tar_file_name in member.name: + filename = tar_file_name + if filename is None: continue file = tarObj.extractfile(member) with open( @@ -176,43 +181,43 @@ def parse_args(mMain=True, add_help=True): parser.add_argument("--use_angle_cls", type=str2bool, default=False) return parser.parse_args() else: - return argparse.Namespace(use_gpu=True, - ir_optim=True, - use_tensorrt=False, - gpu_mem=8000, - image_dir='', - det_algorithm='DB', - det_model_dir=None, - det_limit_side_len=960, - det_limit_type='max', - det_db_thresh=0.3, - det_db_box_thresh=0.5, - det_db_unclip_ratio=2.0, - det_east_score_thresh=0.8, - det_east_cover_thresh=0.1, - det_east_nms_thresh=0.2, - rec_algorithm='CRNN', - rec_model_dir=None, - rec_image_shape="3, 32, 320", - rec_char_type='ch', - rec_batch_num=30, - max_text_length=25, - rec_char_dict_path=None, - use_space_char=True, - drop_score=0.5, - cls_model_dir=None, - cls_image_shape="3, 48, 192", - label_list=['0', '180'], - cls_batch_num=30, - cls_thresh=0.9, - enable_mkldnn=False, - use_zero_copy_run=False, - use_pdserving=False, - lang='ch', - det=True, - rec=True, - use_angle_cls=False - ) + return argparse.Namespace( + use_gpu=True, + ir_optim=True, + use_tensorrt=False, + gpu_mem=8000, + image_dir='', + det_algorithm='DB', + det_model_dir=None, + det_limit_side_len=960, + det_limit_type='max', + det_db_thresh=0.3, + det_db_box_thresh=0.5, + det_db_unclip_ratio=2.0, + det_east_score_thresh=0.8, + det_east_cover_thresh=0.1, + det_east_nms_thresh=0.2, + rec_algorithm='CRNN', + rec_model_dir=None, + rec_image_shape="3, 32, 320", + rec_char_type='ch', + rec_batch_num=30, + max_text_length=25, + rec_char_dict_path=None, + use_space_char=True, + drop_score=0.5, + cls_model_dir=None, + cls_image_shape="3, 48, 192", + label_list=['0', '180'], + cls_batch_num=30, + cls_thresh=0.9, + enable_mkldnn=False, + use_zero_copy_run=False, + use_pdserving=False, + lang='ch', + det=True, + rec=True, + use_angle_cls=False) class PaddleOCR(predict_system.TextSystem): @@ -228,19 +233,21 @@ class PaddleOCR(predict_system.TextSystem): lang = postprocess_params.lang assert lang in model_urls[ 'rec'], 'param lang must in {}, but got {}'.format( - model_urls['rec'].keys(), lang) + model_urls['rec'].keys(), lang) if postprocess_params.rec_char_dict_path is None: postprocess_params.rec_char_dict_path = model_urls['rec'][lang][ 'dict_path'] # init model dir if postprocess_params.det_model_dir is None: - postprocess_params.det_model_dir = os.path.join(BASE_DIR, 'det') + postprocess_params.det_model_dir = os.path.join( + BASE_DIR, '{}/det'.format(VERSION)) if postprocess_params.rec_model_dir is None: postprocess_params.rec_model_dir = os.path.join( - BASE_DIR, 'rec/{}'.format(lang)) + BASE_DIR, '{}/rec/{}'.format(VERSION, lang)) if postprocess_params.cls_model_dir is None: - postprocess_params.cls_model_dir = os.path.join(BASE_DIR, 'cls') + postprocess_params.cls_model_dir = os.path.join( + BASE_DIR, '{}/cls'.format(VERSION)) print(postprocess_params) # download model maybe_download(postprocess_params.det_model_dir, model_urls['det']) diff --git a/ppocr/metrics/cls_metric.py b/ppocr/metrics/cls_metric.py index 03cbe9c80e12f483f3aa5da93a7bf58683596e6b..09817200234dc8d8b5d091ebbe33f07f4aad2cf6 100644 --- a/ppocr/metrics/cls_metric.py +++ b/ppocr/metrics/cls_metric.py @@ -32,9 +32,8 @@ class ClsMetric(object): def get_metric(self): """ - return metircs { - 'acc': 0, - 'norm_edit_dis': 0, + return metrics { + 'acc': 0 } """ acc = self.correct_num / self.all_num diff --git a/ppocr/metrics/det_metric.py b/ppocr/metrics/det_metric.py index 889a8e152254365f9c4d417125e2e642577660b5..0f9e94df42bb8f31ebc79693a01968d441b16faa 100644 --- a/ppocr/metrics/det_metric.py +++ b/ppocr/metrics/det_metric.py @@ -57,7 +57,7 @@ class DetMetric(object): def get_metric(self): """ - return metircs { + return metrics { 'precision': 0, 'recall': 0, 'hmean': 0 diff --git a/ppocr/metrics/rec_metric.py b/ppocr/metrics/rec_metric.py index 98817ad82952bb07a39d594cb6994a5460aff496..bd0f92e0d759204b33b6cb9b261531d61134605e 100644 --- a/ppocr/metrics/rec_metric.py +++ b/ppocr/metrics/rec_metric.py @@ -43,7 +43,7 @@ class RecMetric(object): def get_metric(self): """ - return metircs { + return metrics { 'acc': 0, 'norm_edit_dis': 0, } diff --git a/ppocr/postprocess/db_postprocess.py b/ppocr/postprocess/db_postprocess.py old mode 100644 new mode 100755 index 0be2c12ad4bb9e70708c45d3e3f60dd526dc4e83..16c789dcd7e9740ca8ddf613d0f2567c9af22820 --- a/ppocr/postprocess/db_postprocess.py +++ b/ppocr/postprocess/db_postprocess.py @@ -40,7 +40,7 @@ class DBPostProcess(object): self.max_candidates = max_candidates self.unclip_ratio = unclip_ratio self.min_size = 3 - self.dilation_kernel = None if not use_dilation else [[1, 1], [1, 1]] + self.dilation_kernel = None if not use_dilation else np.array([[1, 1], [1, 1]]) def boxes_from_bitmap(self, pred, _bitmap, dest_width, dest_height): ''' diff --git a/ppocr/utils/dict/en_dict.txt b/ppocr/utils/dict/en_dict.txt new file mode 100644 index 0000000000000000000000000000000000000000..6fbd99f46acca8391a5e86ae546c637399204506 --- /dev/null +++ b/ppocr/utils/dict/en_dict.txt @@ -0,0 +1,63 @@ +0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +a +b +c +d +e +f +g +h +i +j +k +l +m +n +o +p +q +r +s +t +u +v +w +x +y +z +A +B +C +D +E +F +G +H +I +J +K +L +M +N +O +P +Q +R +S +T +U +V +W +X +Y +Z + diff --git a/ppocr/utils/dict/french_dict.txt b/ppocr/utils/dict/french_dict.txt index a74c60ad39a8382da4658b4af3adc2d9b1df317c..e8f657db35bf0b74f779f38a9e3b9b47b007e3c4 100644 --- a/ppocr/utils/dict/french_dict.txt +++ b/ppocr/utils/dict/french_dict.txt @@ -132,4 +132,5 @@ j ³ Å $ -# \ No newline at end of file +# + diff --git a/ppocr/utils/dict/german_dict.txt b/ppocr/utils/dict/german_dict.txt index ba9d472adc5a1c15b860a3cbb934cec97db4f268..af0b01ebc9c5f588b621e318a9d85760cd8f42d9 100644 --- a/ppocr/utils/dict/german_dict.txt +++ b/ppocr/utils/dict/german_dict.txt @@ -123,4 +123,5 @@ z â å æ -é \ No newline at end of file +é + diff --git a/ppocr/utils/dict/japan_dict.txt b/ppocr/utils/dict/japan_dict.txt index 926979bc6b901e915931480010bee4aa720a8cbb..339d4b89e5159a346636641a0814874faa59754a 100644 --- a/ppocr/utils/dict/japan_dict.txt +++ b/ppocr/utils/dict/japan_dict.txt @@ -4395,4 +4395,5 @@ z ｙ ｚ ～ -･ \ No newline at end of file +･ + diff --git a/ppocr/utils/dict/korean_dict.txt b/ppocr/utils/dict/korean_dict.txt index 77ae5c301a1173622b1537d01da34009f2c40a4c..a13899f14dfe3bfc25b34904390c7b1e4ed8674b 100644 --- a/ppocr/utils/dict/korean_dict.txt +++ b/ppocr/utils/dict/korean_dict.txt @@ -179,7 +179,7 @@ z с т я - +​ ’ “ ” @@ -3684,4 +3684,5 @@ z 立 茶 切 -宅 \ No newline at end of file +宅 + diff --git a/ppocr/utils/ic15_dict.txt b/ppocr/utils/ic15_dict.txt index 71043689051fb5a2da516b2e005d1d9b0fdecfb3..474060366f8a2a00c108d5c743821c0a61867cd5 100644 --- a/ppocr/utils/ic15_dict.txt +++ b/ppocr/utils/ic15_dict.txt @@ -33,4 +33,4 @@ v w x y -z +z \ No newline at end of file diff --git a/tools/export_model.py b/tools/export_model.py index cf568884f695f647c50d375d50d168d5ca1ea86e..74357d58ec977bf21ec56d12043c0985bad1f817 100755 --- a/tools/export_model.py +++ b/tools/export_model.py @@ -28,37 +28,16 @@ from ppocr.modeling.architectures import build_model from ppocr.postprocess import build_post_process from ppocr.utils.save_load import init_model from ppocr.utils.logging import get_logger -from tools.program import load_config - - -def parse_args(): - parser = argparse.ArgumentParser() - parser.add_argument("-c", "--config", help="configuration file to use") - parser.add_argument( - "-o", "--output_path", type=str, default='./output/infer/') - return parser.parse_args() - - -class Model(paddle.nn.Layer): - def __init__(self, model): - super(Model, self).__init__() - self.pre_model = model - - # Please modify the 'shape' according to actual needs - @to_static(input_spec=[ - paddle.static.InputSpec( - shape=[None, 3, 640, 640], dtype='float32') - ]) - def forward(self, inputs): - x = self.pre_model(inputs) - return x +from tools.program import load_config, merge_config, ArgsParser def main(): - FLAGS = parse_args() + FLAGS = ArgsParser().parse_args() config = load_config(FLAGS.config) + merge_config(FLAGS.opt) logger = get_logger() # build post process + post_process_class = build_post_process(config['PostProcess'], config['Global']) @@ -71,9 +50,15 @@ def main(): init_model(config, model, logger) model.eval() - model = Model(model) - save_path = '{}/{}'.format(FLAGS.output_path, - config['Architecture']['model_type']) + save_path = '{}/inference'.format(config['Global']['save_inference_dir']) + infer_shape = [3, 32, 100] if config['Architecture'][ + 'model_type'] != "det" else [3, 640, 640] + model = to_static( + model, + input_spec=[ + paddle.static.InputSpec( + shape=[None] + infer_shape, dtype='float32') + ]) paddle.jit.save(model, save_path) logger.info('inference model is saved to {}'.format(save_path)) diff --git a/tools/infer/predict_det.py b/tools/infer/predict_det.py index 5be27339dbae07c8d99fe442f18e64288d831f79..6f98ded8295dabbd5edf05913245e5d94d856689 100755 --- a/tools/infer/predict_det.py +++ b/tools/infer/predict_det.py @@ -63,6 +63,7 @@ class TextDetector(object): postprocess_params["box_thresh"] = args.det_db_box_thresh postprocess_params["max_candidates"] = 1000 postprocess_params["unclip_ratio"] = args.det_db_unclip_ratio + postprocess_params["use_dilation"] = True else: logger.info("unknown det_algorithm:{}".format(self.det_algorithm)) sys.exit(0) @@ -111,7 +112,7 @@ class TextDetector(object): box = self.clip_det_res(box, img_height, img_width) rect_width = int(np.linalg.norm(box[0] - box[1])) rect_height = int(np.linalg.norm(box[0] - box[3])) - if rect_width <= 10 or rect_height <= 10: + if rect_width <= 3 or rect_height <= 3: continue dt_boxes_new.append(box) dt_boxes = np.array(dt_boxes_new) @@ -186,4 +187,4 @@ if __name__ == "__main__": cv2.imwrite(img_path, src_im) logger.info("The visualized image saved in {}".format(img_path)) if count > 1: - logger.info("Avg Time:", total_time / (count - 1)) + logger.info("Avg Time: {}".format(total_time / (count - 1))) diff --git a/tools/infer/utility.py b/tools/infer/utility.py index fabc33dc67265cd304294bab15ecd6d242a3add6..4b06b60b9e25954be7375882b5fb67343312b222 100755 --- a/tools/infer/utility.py +++ b/tools/infer/utility.py @@ -100,8 +100,8 @@ def create_predictor(args, mode, logger): if model_dir is None: logger.info("not find {} model file path {}".format(mode, model_dir)) sys.exit(0) - model_file_path = model_dir + "/model" - params_file_path = model_dir + "/params" + model_file_path = model_dir + "/inference.pdmodel" + params_file_path = model_dir + "/inference.pdiparams" if not os.path.exists(model_file_path): logger.info("not find model file path {}".format(model_file_path)) sys.exit(0) diff --git a/tools/program.py b/tools/program.py index 8e84d30e64fa19a99fea205bca2d08c490b6fd7e..787a59d49b9963421c99b17bd563ddc10a2a601b 100755 --- a/tools/program.py +++ b/tools/program.py @@ -113,7 +113,6 @@ def merge_config(config): global_config.keys(), sub_keys[0]) cur = global_config[sub_keys[0]] for idx, sub_key in enumerate(sub_keys[1:]): - assert (sub_key in cur) if idx == len(sub_keys) - 2: cur[sub_key] = value else: