This repo covers our solution for MICCAI2022 Diabetic Retinopathy Analysis Challenge (DRAC), which solves the task of rating OCTA image quality according to poor, good and excellent.
Download the dataset from the official website: https://drac22.grand-challenge.org/
# 1. Create a conda virtual environment.
conda create -n drac python=3.9 -y
conda activate drac
# 2. Install PyTorch (We use PyTorch 2.0 built under cuda 11.8)
conda install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia
# 3.
pip install timm
pip install pandas
pip install scikit-learn
pip install torchnet
.
├── checkpoints
│ ├── effb2
│ ├── incepv3
│ ├── resnet18
│ └── resnet50d
├── data
│ ├── 1. Original Images
│ │ ├── a. Training Set
│ │ └── b. Testing Set
│ ├── 2. Groundtruths
│ └── LICENSE.txt
├── dataset.py
├── main.py
├── README.md
├── results
│ ├── effb2
│ ├── incepv3
│ ├── resnet18
│ └── resnet50d
├── test_multi.py
└── test.py
python main.py --model resnet50d
python test_multi.py --model resnet50d