diff --git a/doc/doc_ch/FAQ.md b/doc/doc_ch/FAQ.md index eddc1d77..2e294229 100644 --- a/doc/doc_ch/FAQ.md +++ b/doc/doc_ch/FAQ.md @@ -38,15 +38,13 @@ PaddleOCR已完成Windows和Mac系统适配,运行时注意两点:1、在[ 中文数据集,LSVT街景数据集训练数据3w张图片 - 识别: 英文数据集,MJSynth和SynthText合成数据,数据量上千万。 - 中文数据集,LSVT街景数据集根据真值将图crop出来,并进行位置校准,总共30w张图像。此外基于LSVT的语料,合成数据500w。 - + 中文数据集,LSVT街景数据集根据真值将图crop出来,并进行位置校准,总共30w张图像。此外基于LSVT的语料,合成数据500w。 + 其中,公开数据集都是开源的,用户可自行搜索下载,也可参考[中文数据集](./datasets.md),合成数据暂不开源,用户可使用开源合成工具自行合成,可参考的合成工具包括[text_renderer](https://github.com/Sanster/text_renderer)、[SynthText](https://github.com/ankush-me/SynthText)、[TextRecognitionDataGenerator](https://github.com/Belval/TextRecognitionDataGenerator)等。 -10. **使用带TPS的识别模型预测报错** - -报错信息:Input(X) dims[3] and Input(Grid) dims[2] should be equal, but received X dimension[3](320) != Grid dimension[2](100) +10. **使用带TPS的识别模型预测报错** +报错信息:Input(X) dims[3] and Input(Grid) dims[2] should be equal, but received X dimension[3](320) != Grid dimension[2](100) 原因:TPS模块暂时无法支持变长的输入,请设置 --rec_image_shape='3,32,100' --rec_char_type='en' 固定输入shape -11. **自定义字典训练的模型,识别结果出现字典里没出现的字** - +11. **自定义字典训练的模型,识别结果出现字典里没出现的字** 预测时没有设置采用的自定义字典路径。设置方法是在预测时,通过增加输入参数rec_char_dict_path来设置。 diff --git a/doc/doc_ch/inference.md b/doc/doc_ch/inference.md index ea4bc756..68cd3938 100644 --- a/doc/doc_ch/inference.md +++ b/doc/doc_ch/inference.md @@ -127,7 +127,7 @@ python3 tools/infer/predict_det.py --image_dir="./doc/imgs_en/img_10.jpg" --det_ ## 文本识别模型推理 -下面将介绍超轻量中文识别模型推理和基于CTC损失的识别模型推理。**而基于Attention损失的识别模型推理还在调试中**。对于中文文本识别,建议优先选择基于CTC损失的识别模型,实践中也发现基于Attention损失的效果不如基于CTC损失的识别模型。 +下面将介绍超轻量中文识别模型推理、基于CTC损失的识别模型推理和基于Attention损失的识别模型推理。对于中文文本识别,建议优先选择基于CTC损失的识别模型,实践中也发现基于Attention损失的效果不如基于CTC损失的识别模型。此外,如果训练时修改了文本的字典,请参考下面的自定义文本识别字典的推理。 ### 1.超轻量中文识别模型推理 diff --git a/doc/doc_en/FAQ_en.md b/doc/doc_en/FAQ_en.md index cdbc6bf7..f4a4499e 100644 --- a/doc/doc_en/FAQ_en.md +++ b/doc/doc_en/FAQ_en.md @@ -44,8 +44,6 @@ At present, the open source model, dataset and magnitude are as follows: Among them, the public datasets are opensourced, users can search and download by themselves, or refer to [Chinese data set](./datasets_en.md), synthetic data is not opensourced, users can use open-source synthesis tools to synthesize data themselves. Current available synthesis tools include [text_renderer](https://github.com/Sanster/text_renderer), [SynthText](https://github.com/ankush-me/SynthText), [TextRecognitionDataGenerator](https://github.com/Belval/TextRecognitionDataGenerator), etc. -10. **Error in using the model with TPS module for prediction** - -Error message: Input(X) dims[3] and Input(Grid) dims[2] should be equal, but received X dimension[3](108) != Grid dimension[2](100) - +10. **Error in using the model with TPS module for prediction** +Error message: Input(X) dims[3] and Input(Grid) dims[2] should be equal, but received X dimension[3](108) != Grid dimension[2](100) Solution:TPS does not support variable shape. Please set --rec_image_shape='3,32,100' and --rec_char_type='en' diff --git a/tools/eval_utils/eval_det_utils.py b/tools/eval_utils/eval_det_utils.py index 252c9364..ba275eca 100644 --- a/tools/eval_utils/eval_det_utils.py +++ b/tools/eval_utils/eval_det_utils.py @@ -59,7 +59,13 @@ def cal_det_res(exe, config, eval_info_dict): img_list.append(data[ino][0]) ratio_list.append(data[ino][1]) img_name_list.append(data[ino][2]) - img_list = np.concatenate(img_list, axis=0) + try: + img_list = np.concatenate(img_list, axis=0) + except: + err = "concatenate error usually caused by different input image shapes in evaluation or testing.\n \ + Please set \"test_batch_size_per_card\" in main yml as 1\n \ + or add \"test_image_shape: [h, w]\" in reader yml for EvalReader." + raise Exception(err) outs = exe.run(eval_info_dict['program'], \ feed={'image': img_list}, \ fetch_list=eval_info_dict['fetch_varname_list'])