modify vitstr name

c269e1e4 · Topdu · bc5522b6 · c269e1e4 · c269e1e4 · c269e1e4
12 changed file
--- a/configs/rec/rec_mtb_nrtr.yml
+++ b/configs/rec/rec_mtb_nrtr.yml
@@ -49,7 +49,7 @@ Architecture:
 Loss:
-  name: CESmoothingLoss
+  name: CELoss
  smoothing: True
 PostProcess:

--- a/configs/rec/rec_vitstr.yml
+++ b/configs/rec/rec_vitstr.yml
@@ -3,7 +3,7 @@ Global:
  epoch_num: 20
  log_smooth_window: 20
  print_batch_step: 10
-  save_model_dir: ./output/rec/vitstr/
+  save_model_dir: ./output/rec/vitstr_none_ce/
  save_epoch_step: 1
  # evaluation is run every 2000 iterations after the 0th iteration#
  eval_batch_step: [0, 50]
@@ -44,7 +44,7 @@ Architecture:
    name: CTCHead
 Loss:
-  name: CESmoothingLoss
+  name: CELoss
  smoothing: False
  with_all: True

--- a/doc/doc_ch/algorithm_overview.md
+++ b/doc/doc_ch/algorithm_overview.md
@@ -85,7 +85,7 @@
 |SAR|Resnet31| 87.20% | rec_r31_sar | [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_r31_sar_train.tar) |
 |SEED|Aster_Resnet| 85.35% | rec_resnet_stn_bilstm_att | [训练模型](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_resnet_stn_bilstm_att.tar) |
 |SVTR|SVTR-Tiny| 89.25% | rec_svtr_tiny_none_ctc_en | [训练模型](https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/rec_svtr_tiny_none_ctc_en_train.tar) |
-|ViTSTR|ViTSTR| 79.82% | rec_vitstr_none_ctc_en | [训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar) |
+|ViTSTR|ViTSTR| 79.82% | rec_vitstr_none_ce_en | [训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_ce_train.tar) |
 <a name="2"></a>

--- a/doc/doc_ch/algorithm_rec_vitstr.md
+++ b/doc/doc_ch/algorithm_rec_vitstr.md
@@ -27,7 +27,7 @@
 |模型|骨干网络|配置文件|Acc|下载链接|
 | --- | --- | --- | --- | --- |
-|ViTSTR|ViTSTR|[rec_vitstr.yml](../../configs/rec/rec_vitstr.yml)|79.82%|[训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar)|
+|ViTSTR|ViTSTR|[rec_vitstr_none_ce.yml](../../configs/rec/rec_vitstr_none_ce.yml)|79.82%|[训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_ce_train.tar)|
 <a name="2"></a>
 ## 2. 环境配置
@@ -40,7 +40,7 @@
 <a name="3-1"></a>
 ### 3.1 模型训练
-请参考[文本识别训练教程](./recognition.md)。PaddleOCR对代码进行了模块化，训练`ViTSTR`识别模型时需要**更换配置文件**为`ViTSTR`的[配置文件](../../configs/rec/rec_ViTSTR.yml)。
+请参考[文本识别训练教程](./recognition.md)。PaddleOCR对代码进行了模块化，训练`ViTSTR`识别模型时需要**更换配置文件**为`ViTSTR`的[配置文件](../../configs/rec/rec_vitstr_none_ce.yml)。
 #### 启动训练
@@ -48,10 +48,10 @@
 具体地，在完成数据准备后，便可以启动训练，训练命令如下：
 ```shell
 #单卡训练（训练周期长，不建议）
-python3 tools/train.py -c configs/rec/rec_vitstr.yml
+python3 tools/train.py -c configs/rec/rec_vitstr_none_ce.yml
 #多卡训练，通过--gpus参数指定卡号
-python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/rec/rec_vitstr.yml
+python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/rec/rec_vitstr_none_ce.yml
 ```
 <a name="3-2"></a>
@@ -61,7 +61,7 @@ python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs
 ```shell
 # 注意将pretrained_model的路径设置为本地路径。
-python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_vitstr.yml -o Global.pretrained_model=./rec_vitstr_train/best_accuracy
+python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.pretrained_model=./rec_vitstr_none_ce_train/best_accuracy
 ```
 <a name="3-3"></a>
@@ -70,7 +70,7 @@ python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec
 使用如下命令进行单张图片预测：
 ```shell
 # 注意将pretrained_model的路径设置为本地路径。
-python3 tools/infer_rec.py -c configs/rec/rec_vitstr.yml -o Global.infer_img='./doc/imgs_words_en/word_10.png' Global.pretrained_model=./rec_vitstr_train/best_accuracy
+python3 tools/infer_rec.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.infer_img='./doc/imgs_words_en/word_10.png' Global.pretrained_model=./rec_vitstr_none_ce_train/best_accuracy
 # 预测文件夹下所有图像时，可修改infer_img为文件夹，如 Global.infer_img='./doc/imgs_words_en/'。
 ```
@@ -80,15 +80,15 @@ python3 tools/infer_rec.py -c configs/rec/rec_vitstr.yml -o Global.infer_img='./
 <a name="4-1"></a>
 ### 4.1 Python推理
-首先将训练得到best模型，转换成inference model。这里以训练完成的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar) )，可以使用如下命令进行转换：
+首先将训练得到best模型，转换成inference model。这里以训练完成的模型为例（[模型下载地址](https://paddleocr.bj.bcebos.com/rec_vitstr_none_ce_train.tar) )，可以使用如下命令进行转换：
 ```shell
 # 注意将pretrained_model的路径设置为本地路径。
-python3 tools/export_model.py -c configs/rec/rec_vitstr.yml -o Global.pretrained_model=./rec_vitstr_train/best_accuracy Global.save_inference_dir=./inference/rec_vitstr/
+python3 tools/export_model.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.pretrained_model=./rec_vitstr_none_ce_train/best_accuracy Global.save_inference_dir=./inference/rec_vitstr/
 ```
 **注意：**
 - 如果您是在自己的数据集上训练的模型，并且调整了字典文件，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。
- 如果您修改了训练时的输入大小，请修改`tools/export_model.py`文件中的对应NRTR的`infer_shape`。
+- 如果您修改了训练时的输入大小，请修改`tools/export_model.py`文件中的对应ViTSTR的`infer_shape`。
 转换成功后，在目录下有三个文件：
 ```
@@ -110,20 +110,20 @@ python3 tools/infer/predict_rec.py --image_dir='./doc/imgs_words_en/word_10.png'
 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：
 结果如下：
 ```shell
-Predicts of ./doc/imgs_words_en/word_10.png:('pain', 0.9265879392623901)
+Predicts of ./doc/imgs_words_en/word_10.png:('pain', 0.9998350143432617)
 ```
 **注意**：
 - 训练上述模型采用的图像分辨率是[1，224，224]，需要通过参数`rec_image_shape`设置为您训练时的识别图像形状。
 - 在推理时需要设置参数`rec_char_dict_path`指定字典，如果您修改了字典，请修改该参数为您的字典文件。
- 如果您修改了预处理方法，需修改`tools/infer/predict_rec.py`中NRTR的预处理为您的预处理方法。
+- 如果您修改了预处理方法，需修改`tools/infer/predict_rec.py`中ViTSTR的预处理为您的预处理方法。
 <a name="4-2"></a>
 ### 4.2 C++推理部署
-由于C++预处理后处理还未支持NRTR，所以暂未支持
+由于C++预处理后处理还未支持ViTSTR，所以暂未支持
 <a name="4-3"></a>
 ### 4.3 Serving服务化部署
@@ -139,7 +139,7 @@ Predicts of ./doc/imgs_words_en/word_10.png:('pain', 0.9265879392623901)
 ## 5. FAQ
 1. 在`ViTSTR`论文中，使用在ImageNet1k上的预训练权重进行初始化训练，我们在训练未采用预训练权重，最终精度没有变化甚至有所提高。
-2. 我们仅仅复现了`ViTSTR`中的tiny版本，如果有需要使用small、base版本，可直接使用源开源repo中的预训练权重转为Paddle权重即可使用。
+2. 我们仅仅复现了`ViTSTR`中的tiny版本，如果需要使用small、base版本，可将[ViTSTR源repo](https://github.com/roatienza/deep-text-recognition-benchmark) 中的预训练权重转为Paddle权重使用。
 ## 引用

--- a/doc/doc_en/algorithm_overview_en.md
+++ b/doc/doc_en/algorithm_overview_en.md
@@ -84,7 +84,7 @@ Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation r
 |SAR|Resnet31| 87.20% | rec_r31_sar | [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_r31_sar_train.tar) |
 |SEED|Aster_Resnet| 85.35% | rec_resnet_stn_bilstm_att | [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_resnet_stn_bilstm_att.tar) |
 |SVTR|SVTR-Tiny| 89.25% | rec_svtr_tiny_none_ctc_en | [trained model](https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/rec_svtr_tiny_none_ctc_en_train.tar) |
-|ViTSTR|ViTSTR| 79.82% | rec_vitstr_none_ctc_en | [trained model](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar) |
+|ViTSTR|ViTSTR| 79.82% | rec_vitstr_none_ce_en | [trained model](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar) |
 <a name="2"></a>

--- a/doc/doc_en/algorithm_rec_vitstr_en.md
+++ b/doc/doc_en/algorithm_rec_vitstr_en.md
@@ -25,7 +25,7 @@ Using MJSynth and SynthText two text recognition datasets for training, and eval
 |Model|Backbone|config|Acc|Download link|
 | --- | --- | --- | --- | --- |
-|ViTSTR|ViTSTR|[rec_vitstr.yml](../../configs/rec/rec_vitstr.yml)|79.82%|[训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar)|
+|ViTSTR|ViTSTR|[rec_vitstr_none_ce.yml](../../configs/rec/rec_vitstr_none_ce.yml)|79.82%|[训练模型](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar)|
 <a name="2"></a>
 ## 2. Environment
@@ -43,24 +43,24 @@ Specifically, after the data preparation is completed, the training can be start
 ```
 #Single GPU training (long training period, not recommended)
-python3 tools/train.py -c configs/rec/rec_vitstr.yml
+python3 tools/train.py -c configs/rec/rec_vitstr_none_ce.yml
 #Multi GPU training, specify the gpu number through the --gpus parameter
-python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/rec/rec_vitstr.yml
+python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/rec/rec_vitstr_none_ce.yml
 ```
 Evaluation:
 ```
 # GPU evaluation
-python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_vitstr.yml -o Global.pretrained_model={path/to/weights}/best_accuracy
+python3 -m paddle.distributed.launch --gpus '0' tools/eval.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.pretrained_model={path/to/weights}/best_accuracy
 ```
 Prediction:
 ```
 # The configuration file used for prediction must match the training
-python3 tools/infer_rec.py -c configs/rec/rec_vitstr.yml -o Global.infer_img='./doc/imgs_words_en/word_10.png' Global.pretrained_model=./rec_vitstr_train/best_accuracy
+python3 tools/infer_rec.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.infer_img='./doc/imgs_words_en/word_10.png' Global.pretrained_model=./rec_vitstr_none_ce_train/best_accuracy
 ```
 <a name="4"></a>
@@ -71,7 +71,7 @@ python3 tools/infer_rec.py -c configs/rec/rec_vitstr.yml -o Global.infer_img='./
 First, the model saved during the ViTSTR text recognition training process is converted into an inference model. ( [Model download link](https://paddleocr.bj.bcebos.com/rec_vitstr_none_none_train.tar)) ), you can use the following command to convert:
 ```
-python3 tools/export_model.py -c configs/rec/rec_vitstr.yml -o Global.pretrained_model=./rec_vitstr_train/best_accuracy  Global.save_inference_dir=./inference/rec_vitstr
+python3 tools/export_model.py -c configs/rec/rec_vitstr_none_ce.yml -o Global.pretrained_model=./rec_vitstr_none_ce_train/best_accuracy  Global.save_inference_dir=./inference/rec_vitstr
 ```
 **Note:**
@@ -98,7 +98,7 @@ python3 tools/infer/predict_rec.py --image_dir='./doc/imgs_words_en/word_10.png'
 After executing the command, the prediction result (recognized text and score) of the image above is printed to the screen, an example is as follows:
 The result is as follows:
 ```shell
-Predicts of ./doc/imgs_words_en/word_10.png:('pain', 0.9265879392623901)
+Predicts of ./doc/imgs_words_en/word_10.png:('pain', 0.9998350143432617)
 ```
 <a name="4-2"></a>

--- a/ppocr/losses/__init__.py
+++ b/ppocr/losses/__init__.py
@@ -30,7 +30,7 @@ from .det_fce_loss import FCELoss
 from .rec_ctc_loss import CTCLoss
 from .rec_att_loss import AttentionLoss
 from .rec_srn_loss import SRNLoss
-from .rec_ce_smooth_loss import CESmoothingLoss
+from .rec_ce_loss import CELoss
 from .rec_sar_loss import SARLoss
 from .rec_aster_loss import AsterLoss
 from .rec_pren_loss import PRENLoss
@@ -60,9 +60,8 @@ def build_loss(config):
    support_dict = [
        'DBLoss', 'PSELoss', 'EASTLoss', 'SASTLoss', 'FCELoss', 'CTCLoss',
        'ClsLoss', 'AttentionLoss', 'SRNLoss', 'PGLoss', 'CombinedLoss',
-        'CESmoothingLoss', 'TableAttentionLoss', 'SARLoss', 'AsterLoss',
+        'CELoss', 'TableAttentionLoss', 'SARLoss', 'AsterLoss', 'SDMGRLoss',
-        'SDMGRLoss', 'VQASerTokenLayoutLMLoss', 'LossFromOutput', 'PRENLoss',
+        'VQASerTokenLayoutLMLoss', 'LossFromOutput', 'PRENLoss', 'MultiLoss'
-        'MultiLoss'
    ]
    config = copy.deepcopy(config)
    module_name = config.pop('name')

--- a/ppocr/losses/rec_ce_smooth_loss.py
+++ b/ppocr/losses/rec_ce_smooth_loss.py
@@ -3,9 +3,9 @@ from paddle import nn
 import paddle.nn.functional as F
-class CESmoothingLoss(nn.Layer):
+class CELoss(nn.Layer):
    def __init__(self, smoothing=True, with_all=False, **kwargs):
-        super(CESmoothingLoss, self).__init__()
+        super(CELoss, self).__init__()
        self.loss_func = nn.CrossEntropyLoss(reduction='mean', ignore_index=0)
        self.smoothing = smoothing
        self.with_all = with_all

--- a/test_tipc/configs/rec_mtb_nrtr/rec_mtb_nrtr.yml
+++ b/test_tipc/configs/rec_mtb_nrtr/rec_mtb_nrtr.yml
@@ -49,7 +49,7 @@ Architecture:
 Loss:
-  name: CESmoothingLoss
+  name: CELoss
  smoothing: True
 PostProcess:

--- a/test_tipc/configs/rec_mtb_nrtr/train_infer_python.txt
+++ b/test_tipc/configs/rec_mtb_nrtr/train_infer_python.txt
@@ -40,10 +40,10 @@ infer_quant:False
 inference:tools/infer/predict_rec.py --rec_char_dict_path=./ppocr/utils/EN_symbol_dict.txt --rec_image_shape="1,32,100" --rec_algorithm="NRTR"
 --use_gpu:True|False
 --enable_mkldnn:True|False
--cpu_threads:1|6
+--cpu_threads:6
 --rec_batch_num:1|6
 --use_tensorrt:True|False
--precision:fp32|int8
+--precision:fp32
 --rec_model_dir:
 --image_dir:./inference/rec_inference
 --save_log_path:./test/output/

--- a/test_tipc/configs/rec_vitstr/rec_vitstr.yml
+++ b/test_tipc/configs/rec_vitstr/rec_vitstr.yml
@@ -3,7 +3,7 @@ Global:
  epoch_num: 20
  log_smooth_window: 20
  print_batch_step: 10
-  save_model_dir: ./output/rec/vitstr/
+  save_model_dir: ./output/rec/vitstr_none_ce/
  save_epoch_step: 1
  # evaluation is run every 2000 iterations after the 0th iteration#
  eval_batch_step: [0, 2000]
@@ -43,7 +43,7 @@ Architecture:
    name: CTCHead
 Loss:
-  name: CESmoothingLoss
+  name: CELoss
  smoothing: False
  with_all: True

--- a/test_tipc/configs/rec_vitstr/train_infer_python.txt
+++ b/test_tipc/configs/rec_vitstr/train_infer_python.txt
@@ -13,7 +13,7 @@ train_infer_img_dir:./inference/rec_inference
 null:null
 ##
 trainer:norm_train
-norm_train:tools/train.py -c test_tipc/configs/rec_vitstr/rec_vitstr.yml -o
+norm_train:tools/train.py -c test_tipc/configs/rec_vitstr_none_ce/rec_vitstr_none_ce.yml -o
 pact_train:null
 fpgm_train:null
 distill_train:null
@@ -21,21 +21,21 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:tools/eval.py -c test_tipc/configs/rec_vitstr/rec_vitstr.yml -o
+eval:tools/eval.py -c test_tipc/configs/rec_vitstr_none_ce/rec_vitstr_none_ce.yml -o
 null:null
 ##
 ===========================infer_params===========================
 Global.save_inference_dir:./output/
 Global.checkpoints:
-norm_export:tools/export_model.py -c test_tipc/configs/rec_vitstr/rec_vitstr.yml -o
+norm_export:tools/export_model.py -c test_tipc/configs/rec_vitstr_none_ce/rec_vitstr_none_ce.yml -o
 quant_export:null
 fpgm_export:null
 distill_export:null
 export1:null
 export2:null
 ##
-train_model:./inference/rec_vitstr_train/best_accuracy
+train_model:./inference/rec_vitstr_none_ce_train/best_accuracy
-infer_export:tools/export_model.py -c test_tipc/configs/rec_vitstr/rec_vitstr.yml -o
+infer_export:tools/export_model.py -c test_tipc/configs/rec_vitstr_none_ce/rec_vitstr_none_ce.yml -o
 infer_quant:False
 inference:tools/infer/predict_rec.py --rec_char_dict_path=./ppocr/utils/EN_symbol_dict.txt --rec_image_shape="1,224,224" --rec_algorithm="ViTSTR"
 --use_gpu:True|False