STRforIndicLanguages

PyTorch implementation of Scene Text Recognition for Transfer Learning in Indic Languages. We received Best Paper Award for this work at CBDAR'21.

Dependence

• Python3.6.5
• torch==1.2.0
• torchvision==0.4.0
• tensorboard==2.3.0

Train your data

Prepare data

• Follow the instructions in meijieru/crnn.pytorch to create lmdb datasets Use the same step to create separate training and validation synthetic data.

Change parameters and alphabets

Please update the parameters and alphabets according to your requirement.

• Change parameters in the train.py file accordingly
• Input the list of characters (or alphabets) to the train.py file wrt to the language you are training on : The charlist input to the train.py should contain all the characters corresponding to the language present in the dataset, else the program will throw a key error during the training process.

Train

Run mytrain.py --

python3 mytrain.py --trainRoot <\path to the train lmdb dataset> \
--valRoot <\path to the validation lmdb dataset> \
--arch <\architecture:CRNN/STARNET> --lan <\language> --charlist <\path to the character text file> \
--batchSize 32 --nepoch 15 --cuda --expr_dir <\path to the output experiments directory> \
--displayInterval 10 --valInterval 100 --adadelta \ 
--manualSeed 1234 --random_sample --deal_with_lossnan 

Reference

meijieru/crnn.pytorch
Sierkinhane/crnn_chinese_characters_rec