未验证 提交 906558ac 编写于 作者: G Guanghua Yu 提交者: GitHub

update v0.5 release note (#1780)

* update v0.5 release note

* update some comments
上级 8e32932c
......@@ -15,6 +15,7 @@ PaddleDetection模块化地实现了多种主流目标检测算法,提供了
</div>
### 产品动态
- 2020.10.30: PP-YOLO支持矩形图像输入,并新增PACT模型量化策略。
- 2020.10.01: 添加实例分割模型SOLOv2,在Tesla V100上达到38.6 FPS, COCO-val数据集上mask ap达到38.8,预测速度提高24%,mAP提高2.4个百分点。
- 2020.09.30: 发布[移动端检测demo](deploy/android_demo),可直接扫码安装体验。
- 2020.09.21-27: 【目标检测7日打卡课】手把手教你从入门到进阶,深入了解目标检测算法的前世今生。立即加入课程QQ交流群(1136406895)一起学习吧 :)
......@@ -234,7 +235,7 @@ PaddleDetection模块化地实现了多种主流目标检测算法,提供了
## 版本更新
v0.4.0版本已经在`07/2020`发布,增加PP-YOLO, TTFNet, HTC, ACFPN等多个模型,新增BlazeFace人脸关键点检测模型,新增移动端SSDLite系列优化模型,新增GridMask,RandomErasing数据增强方法,新增Matrix NMS和EMA训练,提升易用性,修复已知诸多bug等,详细内容请参考[版本更新文档](docs/CHANGELOG.md)
v0.5.0版本已经在`11/2020`发布,增加SOLOv2等多个模型,新增Android移动端Demo,新增PACT量化策略,新增目标检测全流程和Jetson平台部署等教程,提升易用性,修复已知诸多bug等,详细内容请参考[版本更新文档](docs/CHANGELOG.md)
## 许可证书
本项目的发布受[Apache 2.0 license](LICENSE)许可认证。
......
......@@ -16,6 +16,7 @@ After a long time of industry practice polishing, PaddleDetection has had smooth
### Product dynamic
- 2020.10.30: PP-YOLO support rectangular image input, and add a new PACT quantization strategy for slim。
- 2020.10.01: Added SOLOv2 as an instance segmentation model, which reached 38.6 FPS on a single Tesla V100, 38.8 mask AP on Coco-Val dataset, and inference speed increased by 24% and mAP by 2.4 percentage points.
- 2020.09.30: Released the mobile-side detection demo, and you can directly scan the code for installation experience.
- 2020.09.21-27: [Object detection 7 days of punching class] Hand in hand to teach you from the beginning to the advanced level, in-depth understanding of the object detection algorithm life. Join the course QQ group (1136406895) to study together :)
......@@ -253,7 +254,7 @@ All these models can be get in [Model Zoo](#ModelZoo)
## Updates
v0.4.0 was released at `07/2020`, add PP-YOLO, TTFNet, HTC, ACFPN, etc. And add BlaceFace face landmark detection model, add a series of optimized SSDLite models on mobile side, add data augmentations GridMask and RandomErasing, add Matrix NMS and EMA training, and improved ease of use, fix many known bugs, etc. Please refer to [change log](docs/CHANGELOG.md) for details.
v0.5.0 was released at `11/2020`, add SOLOv2, etc. And add Android mobile demo, add PACT quantization strategy, add object detection full-process documentation and Jetson platform deployment tutorials, and improved ease of use, fix many known bugs, etc. Please refer to [change log](docs/CHANGELOG.md) for details.
## License
......
......@@ -2,6 +2,28 @@
## 最新版本信息
### v0.5.0(11/2020)
- 模型丰富度提升:
- 发布SOLOv2系列模型,其中SOLOv2-Light-R50-VD-DCN-FPN 模型在单卡V100上达到 38.6 FPS,加速24% ,COCO验证集精度达到38.8%, 提升2.4绝对百分点。
- 新增Android移动端检测demo,包括SSD、YOLO系列模型,可直接扫码安装体验。
- 移动端模型优化:
- 新增PACT新量化策略,YOLOv3-Mobilenetv3在COCO数据集上比普通量化相比提升0.7%。
- 易用性提升及功能组件:
- 增强generate_proposal_labels算子功能,规避模型出nan风险。
- 修复deploy下python与C++预测若干问题。
- 统一COCO与VOC数据集下评估流程,支持输出单类AP和P-R曲线。
- PP-YOLO支持矩形输入图像。
- 文档:
- 新增目标检测全流程教程,新增Jetson平台部署教程。
## 历史版本信息
### v0.4.0(07/2020)
- 模型丰富度提升:
- 发布PPYOLO模型,COCO数据集精度达到45.2%,单卡V100预测速度达到72.9 FPS,精度和预测速度优于YOLOv4模型。
......@@ -14,7 +36,7 @@
- 移动端模型优化:
- 新增SSDLite系列优化模型,包括新增GhostNet的Backbone,新增FPN组件等,精度提升0.5%-1.5%。
- 易用性提升功能组件:
- 易用性提升功能组件:
- 新增GridMask, RandomErasing数据增强方法。
- 新增Matrix NMS支持。
- 新增EMA(Exponential Moving Average)训练支持。
......@@ -44,9 +66,6 @@
- 支持迁移学习自动进行shape匹配。
- 优化mask分支评估阶段内存占用。
## 历史版本信息
### v0.2.0(02/2020)
- 新增模型:
- 新增基于CBResNet模型。
......@@ -72,7 +91,7 @@
- 文档:
- 增加详细的数据预处理模块介绍文档以及实现自定义数据Reader文档。
- 增加如何新增算法模型的文档。
- 文档部署到网站: https://paddledetection.readthedocs.io/zh/latest/
- 文档部署到网站: https://paddledetection.readthedocs.io
### 12/2019
- 增加Res2Net模型。
......
......@@ -20,9 +20,25 @@ This document covers how to install PaddleDetection, its dependencies
For general information about PaddleDetection, please see [README.md](https://github.com/PaddlePaddle/PaddleDetection/blob/master/).
## PaddlePaddle
## Install PaddlePaddle
Running PaddleDetection requires PaddlePaddle Fluid v.1.6 and later. please follow the instructions in [installation document](http://www.paddlepaddle.org.cn/).
### Requirements:
- Python2 or Python3 (Only support Python3 for windows)
- CUDA >= 9.0
- cuDNN >= 7.6
- nccl >= 2.1.2
If you need GPU multi-card training, firstly please install NCCL. (Windows does not support nccl).
PaddleDetection depends on PaddlePaddle version relationship:
| PaddleDetection version | PaddlePaddle version | tips |
| :----------------: | :---------------: | :-------: |
| v0.3 | >=1.7 | -- |
| v0.4 | >= 1.8.4 | PP-YOLO依赖1.8.4 |
| v0.5 | >= 1.8.4 | Most models can run with >= 1.8.4, Cascade R-CNN and SOLOv2 depend on 2.0.0.rc |
If you want install paddlepaddle, please follow the instructions in [installation document](http://www.paddlepaddle.org.cn/).
Please make sure your PaddlePaddle installation was successful and the version
of your PaddlePaddle is not lower than required. Verify with the following commands.
......@@ -36,12 +52,6 @@ of your PaddlePaddle is not lower than required. Verify with the following comma
python -c "import paddle; print(paddle.__version__)"
```
### Requirements:
- Python2 or Python3 (Only support Python3 for windows)
- CUDA >= 8.0
- cuDNN >= 5.0
- nccl >= 2.1.2
## Other Dependencies
......
......@@ -34,6 +34,7 @@ PaddleDetection 依赖 PaddlePaddle 版本关系:
| :----------------: | :---------------: | :-------: |
| v0.3 | >=1.7 | -- |
| v0.4 | >= 1.8.4 | PP-YOLO依赖1.8.4 |
| v0.5 | >= 1.8.4 | 大部分模型>=1.8.4即可运行,Cascade R-CNN系列模型与SOLOv2依赖2.0.0.rc版本 |
```
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册