Skip to content

Latest commit

 

History

History
141 lines (87 loc) · 3.58 KB

README.md

File metadata and controls

141 lines (87 loc) · 3.58 KB

Convolutional Recurrent Neural Network + CTCLoss

  • The images folder contain all the cropped images and preprocessing descriptions
  • train.ipynb contains the scripts. Currently its a bit unreadable... will refactor later.

TODOs

  • recover around 1000 images!!! wasted during preprocessing
  • this was trained using raw images... should try training it using processed images(they look better in quality)

Run demo

  • Use the weights expr/FinalnetCRNN_249_105.pth

  • Run demo

    python demo.py -m path/to/model/expr/FinalnetCRNN_249_105.pth -i demo/উত্তরাধিকার_82.jpg

Expected output

উ-----তত্্্তররাাধিিককাাারর => উত্তরাধিকার

Train your data

Prepare data

Folder mode

  1. Put your images in a folder and organize your images in the following format:

    label_number.jpg

    For example:

    • English
    hi_0.jpg hello_1.jpg English_2.jpg English_3.jpg E n g l i s h_4.jpg...
    • Bengali
         উত্তরাধিকার_82.jpg দেশপ্রেমের_70.jpg দেশের_13.jpg ধূলিকণায়_16.jpg বিশাল_6.jpg ...

    The numbers are used to distinguish the same labels.

  2. Run the create_dataset.py in tool folder by

    python tool/create_dataset.py --out lmdb/data/output/path --folder path/to/folder

  3. Use the same step to create train and val data.

  4. The advantage of the folder mode is that it's convenient! But due to some illegal character can't be in the path

    So the disadvantage of the folder mode is that it's labels are limited.

File mode

  1. Your data file should like

    absolute/path/to/image/উত্তরাধিকার_82.jpg
     উত্তরাধিকার_82
absolute/path/to/image/দেশপ্রেমের_70.jpg
     দেশপ্রেমের_70
absolute/path/to/image/দেশের_13.jpg
     দেশের_13
absolute/path/to/image/ধূলিকণায়_16.jpg
     ধূলিকণায়
absolute/path/to/image/বিশাল_6.jpg
     বিশাল_6
absolute/path/to/image/xxx.jpg
label of xxx.jpg
.
.
.

    DO REMEMBER:

    1. It must be the absolute path to image.
    2. The first line can't be empty.
    3. There are no blank line between two data.

  2. Run the create_dataset.py in tool folder by

    python tool/create_dataset.py --out lmdb/data/output/path --file path/to/file

  3. Use the same step to create train and val data.

Change parameters and alphabets

Parameters and alphabets can't always be the same in different situation.

  • Change parameters

    Your can see the params.py in detail.

  • Change alphabets

    Please put all the alphabets appeared in your labels to alphabets.py , or the program will throw error during training process.

Train

Run train.py by

python train.py --trainroot path/to/train/dataset --valroot path/to/val/dataset

Dependency

  • Python3.6.5
  • torch==1.2.0
  • torchvision==0.4.0

Reference

meijieru/crnn.pytorch

Sierkinhane/crnn_chinese_characters_rec

Holmeyoung/crnn-pytorch