The official repository. Please cite as
@InProceedings{Nahon_2023_ICCV,
author = {Nahon, R\'emi and Nguyen, Van-Tam and Tartaglione, Enzo},
title = {Mining bias-target Alignment from Voronoi Cells},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2023},
pages = {4946-4955}
}
This is an example implementation of : Mining bias-target Alignment from Voronoi Cells.
This repository is tailored for Biased-MNIST but can be adapted to other datasets.
If you want to work with conda :
conda create -n voronoi_cells_bias_alignment python=3.9
conda activate voronoi_cells_bias_alignment
pip3 install -r pip_requirements.txt
If not, simply install the requirement with :
pip3 install -r pip_requirements.txt
-
In a terminal, get inside the voronoi_cells_bias_alignment folder
-
Run : python3 main.py
-
--dev : to specify the device you want to work on (ex: cpu or cuda:0)
-
--rho : to select a level of digit-color correlation in Biased-MNIST (ex : 0.997)
-
other extra arguments detailed in utils/configs.py