PyTorch re-implementation of Real-time Scene Text Detection with Differentiable Binarization

1. We try the network on Manga109 dataset for manga text detection
2. Try MobileNetv2 and MobileNetV3(Large) as backbone for dbnet.
3. Concatenate a Unet after FPN head

Note: Manga109 dataset is from University of Tokyo, and we are not allowed to public the dataset. If needed, please request permission from http://www.manga109.org/

Mainly reimplement and inherit from :

• https://github.com/SURFZJY/Real-time-Text-Detection-DBNet

Also thanks to these project:

• https://github.com/WenmuZhou/PAN.pytorch
• https://github.com/d-li14/mobilenetv2.pytorch
• https://github.com/d-li14/mobilenetv3.pytorch

The features are summarized blow:

• Use resnet18/resnet50/shufflenetV2/mobilenetV2andV3 as backbone.
• Use Unet as afterburner.

1. Installation
2. Download
3. Train
4. Predict
5. Eval
6. Demo

1. pytorch 1.1.0

1. Go to manga process to generate txt list file for training/testing data

2. modify config.json

3. run

python train.py

1. run

python predict.py

run

python eval.py