https://github.com/RPIDIAL/SAMix
MICCAI 2023 - Spectral Adversarial MixUp for Few-Shot Unsupervised Domain Adaptation
https://github.com/RPIDIAL/SAMix
Last synced: 4 months ago
JSON representation
MICCAI 2023 - Spectral Adversarial MixUp for Few-Shot Unsupervised Domain Adaptation
- Host: GitHub
- URL: https://github.com/RPIDIAL/SAMix
- Owner: RPIDIAL
- Created: 2023-07-20T01:41:14.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-05-30T20:25:37.000Z (11 months ago)
- Last Synced: 2024-08-24T17:26:21.751Z (8 months ago)
- Language: Python
- Size: 895 KB
- Stars: 7
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- Awesome-Mixup - [Code
README
# SAMix -- Spectral Adversarial MixUp for Few-Shot UDA
[-blue.svg)](https://github.com/996icu/996.ICU/blob/master/LICENSE)
[](https://996.icu)
This repository includes open-source codes, detailed experimental results and full references of our __MICCAI 2023__ paper [*Spectral Adversarial MixUp for Few-Shot Unsupervised Domain Adaptation*](https://arxiv.org/abs/2309.01207).
## Overview
The figure above summarizes our algorithm comparisons framework, *Spectral Adversarial MixUp*.
- Our method first computes [DoDiSS map](https://github.com/DIAL-RPI/Spectral-Adversarial-Data-Augmentation/tree/main/AAAIcodeSubmissoin__model_sensitivity_map), which is a variant of our [*AAAI-23 paper*](https://ojs.aaai.org/index.php/AAAI/article/view/26328).
- Then, we run [SAMix](https://github.com/DIAL-RPI/Spectral-Adversarial-Data-Augmentation/tree/main/AAAIcodeSubmissoin__SADA) to generate the augmented data and train the model with the ERM loss, normal UDA loss and the Jenssen-Shannon divergence regularization.
## Prerequisites
- Python 3.8
- PyTorch 1.8.1+
- A computing device with GPU
## Getting started
### Installation
- (Not necessary) Install [Anaconda3](https://www.anaconda.com/products/distribution)
- Install [CUDA 11.0+](https://developer.nvidia.com/cuda-11.0-download-archive)
- Install [PyTorch 1.12.1+](http://pytorch.org/)
Noted that our code is tested based on [PyTorch 1.12.1](http://pytorch.org/)
### Dataset & Preparation
The two medical image datasets used in our work are publicly available.
The __Fundus__ datasets include the [REFUGE](https://refuge.grand-challenge.org/) and the [RIM](https://github.com/miag-ull/rim-one-dl). We follow the [DoFE](https://arxiv.org/pdf/2010.06208.pdf) to preprocess the datasets.
The __Camelyon__ dataset is a subset from the [WILDS](https://wilds.stanford.edu/datasets/). This dataset is already preprocessed by the challenge.
### Measure the sensitivity map of an ERM model
Train and evaluate an ERM model by
```bash
python ./AAAIcodeSubmissoin__model_sensitivity_map/train_ERM.py
```
The model will be saved in `./AAAIcodeSubmissoin__model_sensitivity_map/save_dir`.
Measure the DoDiSS map of an ERM model by
```bash
python ./AAAIcodeSubmissoin__model_sensitivity_map/model_sensitivity_map.py
```
### Train model with Spectral Adversarial Data Mixup (SAMix)
Train and evaluate the models with SADA by
```bash
python ./AAAIcodeSubmissoin__SADA/train_SADA.py
```
The key __SAMix data augmentation module__ is in
```
./AAAIcodeSubmissoin__SADA/sada.py
```
Augmentation settings for all datasets:
- `--epsilon` iteration of the checkpoint to load. #Default: 0.2
- `--step_size` step size of the adversarial attack on the amplitude spectrum. #Default: 0.08
- `--num_steps` batch size of the attack steps. #Default: 5
- `--tau` settings for the early stop acceleration. #Default: 1
- `--random_init` if or not initializing amplitude spertrums with random perturbations. #Default: True
- `--randominit_type` type of random init type. #Default: 'normal'
- `--criterion` loss functions to attack. #Default: 'ce', choices=['ce', 'kl']
- `--direction` neg: standard attack || pos:inverse attack. #Default: neg, choices=['pos','neg']
## Citation
Please cite these papers in your publications if it helps your research:
```bibtex
@article{zhang2022spectral,
title={Spectral Adversarial MixUp for Few-Shot Unsupervised Domain Adaptation},
author={Zhang, Jiajin and Chao, Hanqing and Dhurandhar, Amit and Chen, Pin-Yu and Tajer, Ali and Xu, Yangyang and Yan, Pingkun},
journal={https://arxiv.org/abs/2309.01207},
year={2023}
}
```
## Acknowledgement
We would like to thank the authors we cited in our paper for sharing their codes.