Replies: 8 comments
-
之所以自定义字典,是因为700多个中字符没有包括在默认的中文字典中,所以重新定义了字典,使用的ch_PP-OCRv3_rec_train预训练模型做的fine-tune |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
你自定义字典,参数不用加吗 |
Beta Was this translation helpful? Give feedback.
-
我用的是ch_PP-OCRv3_rec_distillation.yml配置文件,只修改了字典加载路径,数据加载路径,还有显卡的配置,其他没有修改过 |
Beta Was this translation helpful? Give feedback.
-
我是说你跑预测的时候,不用加字典路径吗 |
Beta Was this translation helpful? Give feedback.
-
![]() 跑预测也添加字典路径了,和训练时的字典一致 |
Beta Was this translation helpful? Give feedback.
-
我也遇到相似的问题,我只在字典里面新增了一个字,但是我识别的时候,比如“20号”识别出来“20”,不知道如何解决,希望有大佬解答 |
Beta Was this translation helpful? Give feedback.
-
训练的时候使用预训练模型吗?还是完全冲头训练? |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
版本是PPOCR2.7
自定义了字典,字典内容是700多个中文字符和0-9,训练和推理都用了自定义字典,train和validation的准确率都没有什么问题,测试时也没有报错。
但是运行一下指令:
python tools/infer/predict_system.py --image_dir="./train_data/icdar2015/text_localization/test_data/" --det_model_dir="./inference_model/ch_PP-OCRv3_det_infer/Student2/" --rec_model_dir="./inference_model/ch_PP-OCRv3_rec_distillation_inference_adc/Teacher/
只能输出数字,无法输出中文
按照#7960的方法修改后,还是一样的问题。
想问一下有什么解决方法或者可以尝试的方向吗?
谢谢~
Beta Was this translation helpful? Give feedback.
All reactions