This repository contains the source code for the experiments of the article

"OLÉ: Orthogonal Low-rank Embedding, A Plug and Play Geometric Loss for Deep Learning" 
José Lezama, Qiang Qiu, Pablo Musé and Guillermo Sapiro, CVPR 2018

https://arxiv.org/abs/1712.01727

If you find this work useful in your research, please consider citing:

@inproceedings{Lezama2018OLE,
title={OL\'E: Orthogonal Low-rank Embedding, A Plug and Play Geometric Loss for Deep Learning},
author={Lezama, Jos\'e and Qiu, Qiang and Mus\'e, Pablo and Sapiro, Guillermo},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
year={2018}
}

STL-10-Pytorch Contains experiments using small training data on STL-10 database. (This is the simplest experiment to run, I recomend starting here)

Cifar10-Caffe Contains experiments on CIFAR10 using Caffe and a VGG-16 architecture

Facescrub500-Caffe Contains experiments on face dataset Facescrub500

Cifar10-Pytorch Contains experiments on CIFAR10 and CIFAR100 for various architectures.

Download dataset used in the paper here