Sample Selection via Contrastive Fragmentation for Noisy Label Regression

This repository contains the official PyTorch implementation for our ICCV2021 paper.

• Chris Dongjoo Kim*, Sangwoo Moon*, Jihwan Moon, Dongyeon Woo, Gunhee Kim. Sample Selection via Contrastive Fragmentation for Noisy Label Regression. In NeurIPS, 2024 (* equal contribution).

[Paper Link]

System Dependencies

• Python >= 3.9
• CUDA >= 9.0 supported GPU

Installation

Using virtual env is recommended.

# create conda env with python=3.9
conda create -n {ENV_NAME} python=3.9

conda activate {ENV_NAME}

# install required version of torch and torchvision
pip install -f https://download.pytorch.org/whl/torch_stable.html torch==1.13.1+cu116 torchvision==0.14.1+cu116

# install other packages
pip install -r requirements.txt

Preparation

Log directory set-up

Create checkpoints directory for logging.

$ ln -s [log directory path] checkpoints

Data set-up:

1. Download data.zip from here.
2. Decompress into ./data.

Run

Specify parameters in config yaml, episodes yaml files.

python main.py --log-dir [log directory path] --config [config file path] --episode [episode file path] --override "|" --random_seed [seed]

Example of ConFrag IMDB-clean-bal run

python main.py --config=configs/imdb_fragment.yaml --episode=episodes/imdb-split4.yaml --log-dir=checkpoints/imdb/dfragment/imdb_clean_bal --random_seed=[seed]

Citation

The code and dataset are free to use for academic purposes only. If you use any of the material in this repository as part of your work, we ask you to cite:

@inproceedings{kim-NeurIPS-2024,
    author    = {Chris Dongjoo Kim and Sangwoo Moon and Jihwan Moon and Dongyeon Woo and Gunhee Kim},
    title     = "{Sample Selection via Contrastive Fragmentation for Noisy Label Regression}"
    booktitle = {NeurIPS},
    year      = 2024
}