This software implements the Convolutional Recurrent Neural Network (CRNN) in pytorch. Origin software could be found in crnn
A demo program can be found in src/demo.py. Before running the demo, download a pretrained model
from Baidu Netdisk or Dropbox.
This pretrained model is converted from auther offered one by tool.
Put the downloaded model file crnn.pth into directory data/. Then launch the demo by:
python demo.py
The demo reads an example image and recognizes its text content.
Expected output: loading pretrained model from ./data/crnn.pth a-----v--a-i-l-a-bb-l-ee-- => available
- warp_ctc_pytorch
- lmdb
- Construct dataset following origin guide. For training with variable length, please sort the image according to the text length.
python crnn_main.py [--param val]. Explorecrnn_main.pyfor details.
