This is an implementation of “Deep relational reasoning graph network for arbitrary shape text detection”.
python 3.7;
PyTorch 1.2.0;
Numpy >=1.16;
CUDA 10.1;
GCC >=9.0;
NVIDIA GPU(with 10G or larger GPU memory for inference);
cd ./csrc and make
cd ./nmslib/lanms and make
Note: download the data and put it under the data file
cd tool
sh train_CTW1500.sh # run or other shell script
you should modify the relevant training parameters according to the environment, such as gpu_id and input_size:
#!/bin/bash
cd ../
CUDA_LAUNCH_BLOCKING=1 python train_textsnake.py --exp_name Ctw1500 --max_epoch 600 --batch_size 6 --gpu 0 --input_size 640 --optim SGD --lr 0.001 --start_epoch 0 --viz --net vgg --resume pretrained/mlt2017_pretain/textsnake_vgg_100.pth
First, you can modify the relevant parameters in the config.py and option.py
python eval_TextGraph.py # Testing single round model
or
python batch_eval.py # Testing multi round models
@InProceedings{Zhang_2020_CVPR, author = {Zhang, Shi-Xue and Zhu, Xiaobin and Hou, Jie-Bo and Liu, Chang and Yang, Chun and Wang, Hongfa and Yin, Xu-Cheng}, title = {Deep Relational Reasoning Graph Network for Arbitrary Shape Text Detection}, booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2020} }
This project is licensed under the MIT License - see the LICENSE.md file for details