Multi-label X-ray Imagery Classification via Bottom-up Attention and Meta Fusion

This repository is the official PyTorch implementation of paper Multi-label X-ray Imagery Classification via Bottom-up Attention and Meta Fusion. (The work has been accepted by ACCV 2020)

Main Requirements

• Python3
• torch == 1.3.1
• torchvision == 0.4.2

Pretrain model

Baidu Cloud password: wf3d

download the pretrain model and put it into the test_offline folder.

Usage

Train

1. get into the main folder:

   cd main/

2. train the attention task (Stage I) as follows:

   sh run_train_bbox_insight.sh

3. train the multi-label classification task (Stage II) as follws:

   sh run_train_attention_multi_label.sh

Inference

1. get into the test_offline folder:

   cd ../test_offline/

2. get inference results:

   python3 test_on_cpu_total_merge.py 

3. Calculate the mAP & FP & FN & TP & TN:

   python3 calc_pre_recall_merge.py

Citation

If you find this work is helpful, please cite this paper.

@inproceedings{Hu2020Xray,
  author    = {Benyi Hu and Chi Zhang and Le Wang and
               Qilin Zhang and Yuehu Liu},
  title     = {Multi-label X-ray Imagery Classification via Bottom-up Attention and Meta Fusion},
  booktitle = {15th Asian Conference on Computer Vision, ({ACCV})},
  year      = {2020},
}