GatingAno

This is an official pytorch implementation of 'Anomaly Detection via Gating Highway Connection for Retinal Fundus Images'.

Requirements

• numpy>=1.17.0
• scipy>=1.5.2
• Pillow>=8.2.0
• pytorch>=1.7.1
• torchvision>=0.8.2
• tqdm>=4.59.0
• scikit-learn>= 0.24.2
• scikit-image>=0.17.2

Datasets

The proposed method is evaluated on two publicly-available datasets, i.e.

• IDRiD
• ADAM

Usage

The proposed GatingAno method is trained through two steps:

• Data Preparation

  Generate the list of HOG image and Patches :

  python3 data_find.py \
  --dataset ['IDRiD'/'IDRiDc'/'ADAM'/'ADAMc'] \
  --path {data dir}

  For example, to generate the image-level label of IDRiD dataset, you can run python3 data_find.py --dataset 'IDRiDc' --path './dataset/'

  And then you will get lists containing images and corresponding labels in './label/IDRiDc/'.

• Training and testing model

  For example, to train pixel-level anomaly detection task on ADAM dataset, you can run

   python3 main.py \
   --dataset 'ADAM' \
   --datadir './labels/ADAM/' \
   --lr 1e-3 \
   --level 'pixel' ;