CRNN_Chinese_Characters_Rec

Characters Recognition

A Chinese characters recognition repository based on convolutional recurrent networks.

Performance

Recognize characters in pictures

Dev Environments

1. WIN 10 or Ubuntu 16.04
2. PyTorch 1.2.0 (may fix ctc loss) with cuda 10.0
3. yaml
4. easydict
5. tensorboardX

Data

Synthetic Chinese String Dataset

1. Download the dataset in here

2. Edit lib/config/360CC_config.yaml DATA:ROOT to you image path

   DATASET:
     ROOT: 'to/your/images/path'
   

3. Put char_std_5990.txt in lib/dataset/txt/

4. Download the labels in here (password: w877)

5. And put train.txt and test.txt in lib/dataset/txt/

   eg. test.txt

   20456343_4045240981.jpg 89 201 241 178 19 94 19 22 26 656
   20457281_3395886438.jpg 120 1061 2 376 78 249 272 272 120 1061
   ...
   

   Your own data

6. Edit lib/config/OWN_config.yaml DATA:ROOT to you image path

   DATASET:
     ROOT: 'to/your/images/path'
   

7. And put your train_own.txt and test_own.txt in lib/dataset/txt/

   eg. test_own.txt

   20456343_4045240981.jpg 你好啊！祖国！
   20457281_3395886438.jpg 晚安啊！世界！
   ...
   

   note: fixed-length supported. yet you can modify dataloader to support random length.

Train

   [run] python train.py --cfg lib/config/360CC_config.yaml

loss curve

   [run] cd output/360CC/crnn/xxxx-xx-xx-xx-xx/
   [run] tensorboard --logdir log

loss overview(first epoch)

Demo

   [run] python demo.py --image_path images/test.png --checkpoints output/checkpoints/mixed_second_finetune_acc_97P7.pth

References

• https://github.com/meijieru/crnn.pytorch
• https://github.com/HRNet