Code of Paper: A Efficient Graph-based Framework for Multi-Organ Histology Image Classification
Extended version of our BMVC 2021 paper: HAT-Net: A Hierarchical Transformer Graph Neural Network for Grading of Colorectal Cancer Histology Images
Our code is partially inspired by CGC-Net, shout out to them.
Some of our installed packages:
- torch-geometric 1.7.2
- torch 1.8.2
- scikit-image 0.17.2
- scikit-learn 1.0
- opencv-python 4.5.3.56
- numpy 1.20.3
- matplotlib 3.4.2
- CRC dataset: a colorectal cancer grading dataset:
- Extended CRC dataset: a colorectal cancer grading dataset.
- UZH dataset: a prostate cancer grading dataset.
- BACH dataset: a breast cancer classification dataset.
python -u train_GIN_Hierarchical.py --norm_adj --batch-size=20 --step_size=15 --jk --cv 3 --num_eval 5 --visualization --epochs 20 --stage 2 3 --depth 6
tested:
python -u train_GIN_Hierarchical.py --norm_adj --batch-size=20 --step_size=15 --jk --cv 3 --num_eval 5 --visualization --lr 0.0003 --epochs 20 --gamma 0.33 --stage 2 3 --depth 1 --load_data_list
or(has not been tested):
python -u train_GIN_Hierarchical.py --norm_adj --batch-size=20 --step_size=15 --jk --cv 3 --num_eval 5 --visualization --lr 0.0003 --epochs 20 --gamma 0.33 --stage 2 3 --depth 1
eval:
python -u eval.py --norm_adj --batch-size=10 --step_size=10 --jk --depth 1 --stage 2 3 --cv 1 --weight_file /data/by/tmp/HGIN/output/result/nuclei_soft-assign_l3x1_ar10_h20_o20_fca_%1fuse_adj0.4_sr1_d0.2_jkknn_cv2/Monday_24_May_2021_09h_50m_38s/model_best.pth.tar
visulization:
python -u train_GIN_Hierarchical.py --norm_adj --batch-size=20 --step_size=15 --jk --cv 3 --num_eval 5 --visualization --lr 0.0003 --epochs 20 --gamma 0.33 --stage 2 3 --depth 1
If you used our code for your research, please cite:
@inproceedings{su2021,
title="HAT-Net: A Hierarchical Transformer Graph Neural Network for Grading of Colorectal Cancer Histology Images",
author={Yihan {Su} and Yu {Bai} and Bo {Zhang} and Zheng {Zhang} and Wendong {Wang}},
booktitle={The British Machine Vision Conference (BMVC)},
year={2021}
}