FYI: Flip Your Images for Dataset Distillation

This repository is an official implementation of the paper "FYI: Flip Your Images for Dataset Distillation" using PyTorch.

For detailed information, please refer to our project site or paper.

Requirements

Please install packages in the requirements.

Getting started

• Prepare the dataset in /dataset directory for experiments on Tiny-ImageNet and ImageNet subset.
• Run the provided scripts as below. Modify the scripts for different settings.

DC, DSA, and DM

cd DC_DSA_DM

# DM for 10 IPC on CIFAR-10
./scripts/dm.sh

# DSA for 10 IPC on CIFAR-10
./scripts/dsa.sh

MTT

cd MTT

# Generating teacher trajectories on CIFAR-10 using ZCA whitening
./scripts/buffer.sh

# MTT for 1 IPC on CIFAR-10
./scripts/distill.sh

Acknowledgement

This implementation is built on DC/DSA/DM and MTT. We also applied our method to IDC and FTD in our paper. We thank the authors and contributors of these projects.

Citation

If you find our work useful for your research, please cite our paper:

@inproceedings{son2024fyi,
  title={{FYI}: Flip Your Images for Dataset Distillation},
  author={Son, Byunggwan and Oh, Youngmin and Baek, Donghyeon and Ham, Bumsub},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  year={2024}
}