Code for LSD-C: Linearly Separable Deep Clusters

by Sylvestre-Alvise Rebuffi*, Sebastien Ehrhardt*, Kai Han*, Andrea Vedaldi, Andrew Zisserman

Dependencies

All dependencies are included in environment.yml. To install the environment, please run:

conda env create -f environment.yml

Then, you can activate the installed environment by running:

conda activate lsd_c

Downloading the pretrained RotNet on CIFAR 10

The pretrained initialization network can be downloaded with the following link:

https://www.dropbox.com/s/4c6jw6caz3tsoe0/RotNet_cifar10.pt?dl=0

Running our clustering method on CIFAR 10

For the kNN labeling method, please run:

python cifar10_clustering.py 

For the cosine labeling method, please run:

python cifar10_clustering.py --similarity_type cosine --hyperparam 0.9

For the SNE labeling method, please run:

python cifar10_clustering.py --similarity_type SNE --hyperparam 0.01

Citation

If this work is helpful for your research, please cite our paper.

@article{rebuffi2020lsdc,
author    = {Sylvestre-Alvise Rebuffi and Sebastien Ehrhardt and Kai Han and Andrea Vedaldi and Andrew Zisserman},
title     = {LSD-C: Linearly Separable Deep Clusters},
journal = {arXiv},
year      = {2020}
}

Acknowledgments

This work is supported by the EPSRC Programme Grant Seebibyte EP/M013774/1, Mathworks/DTA DFR02620, and ERC IDIU-638009.