This is an official implementation of the paper,
Fine-Grained Self-Supervised Learning with Jigsaw Puzzles for Medical Image Classification
Wongi Park, Jongbin Ryu.
(Comming soon..),
- Conda environment
: Ubuntu 18.04 CUDA-10.1 (10.2) with Pytorch==1.13.0, Torchvision==0.6.0 (python 3.8).
# Create Environment
conda create -n fgssl python=3.8
conda activate fgssl
# Install pytorch, torchvision, cudatoolkit
conda install pytorch==1.13.0 torchvision==0.6.0 cudatoolkit=10.1 (10.2) -c pytorch
- How to get dataset?
- Directory structure of our project
- Directory
- run.sh : shell Script Version (pretraining, training, inference)
- main.py : Main Execution (Args, Setting)
- dataset : Augmentation, DataLoader
- ...
- learner : training, inference
- ...
- utils : Metrics, Jigsaw Setting (Metric, jigsaw Patch Setting)
- ...
(1) Jigsaw patch (Paper / Code)
(2) BarlowTwins (Paper / Code)
(3) progressive Learning (Paper / Code)
python main.py --seed 0 --model pmg --patches 32 16 8 --dataset aptos --gpu_ids [gpu_idx]
The results will be automatically saved in ./workspace/[model name]/[Workspace Name].
@article = {
title = {Learning a Mixture of Granularity-Specific Experts for Fine-Grained Categorization},
author = {Wongi Park},
journal = {GitHub},
url = {https://github.com/kalelpark/FG-SSL},
year = {2022},
}