CTPN + CRNN Pytorch Chinese OCR

Chinese characters recognition repository based on CTPN， CRNN

Performance

recognize Chinese Characters from image, here shows some results

Environments

1. WIN 10 or Ubuntu 16.04
2. PyTorch 1.2.0 (may fix ctc loss) with cuda 10.0，recommend 1.6.0
3. yaml
4. easydict
5. tensorboardX
6. opencv
7. pillow
8. numpy

How to use?

[run] python ctpn_crnn.py --image_path images/test_1.png (or change to your image path)

after run above, you could fine result from folder

output/result/result_0.png (or result_1.png depending on how many image you had run)

File instructions

• crnn (folder)
  • include all CRNN scripts
• ctpn_pytorch (folder)
  • include all CTPN scripts
• font (folder)
  • include some fonts support Chinese
• images (folder)
  • include some test images
• output (folder)
  • chechkpoints (folder)
    • include CRNN checkpoint
  • result (folder)
    • include test output
• Download CRNN pretrained weights here (password: kze0)
  • copy to output/checkpoints/
• Download CTPN pretrained weights here (password: 5bpi)
  • copy to ctpn_pytorch/weights/

How to train & Data set

Train CRNN model

1. Data set for CRNN training

   • 3.6 Million Synthetic Chinese String，download from here , (password: auwl)

2. Edit crnn/lib/config/360cc_config.yaml DATA:ROOT to your image path

   • DATASET:360CC
     ROOT: "your image path"
     

3. Download labels (password: fwup)

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

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
   ......
   

   Or your own data

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

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

   2. And put your train_own.txt and test_own.txt in crnn/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 training is supported. yet you can modify dataloader to support random length training.

6. Train

   [run] cd crnn
   [run] python train.py --cfg lib/config/360CC_config.yaml
   or
   [run] python train.py --cfg lib/config/OWN_config.yaml
   

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

7. Test

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

Train CTPN model

1. Data set for CTPN training，download here (password: ffoq)

   this dataset is VOC2007 format

   --VOC2007
    -- Annotations
    -- ImageSets
    -- JPEGImages
   

2. Edit ctpn_pytorch/ctpn/config.py

   img_dir = 'VOC2007/JPEGImages/'
   label_dir = 'VOC2007/Annotations/'
   

3. Train

   cd ctpn_pytorch
   python train.py
   

4. Test

   cd ctpn_pytorch
   python predict.py
   

References

https://github.com/xiaofengShi/CHINESE-OCR

https://github.com/CrazySummerday/ctpn.pytorch