BarlowTwins

<p align="center"> <img width="500" alt="Screen Shot 2021-04-29 at 6 26 48 AM" src="https://user-images.githubusercontent.com/14848164/120419539-b0fab900-c330-11eb-8536-126ce6ce7b85.png"> </p>

This project is a Pytorch implementation from scratch of the paper Barlow Twins: Self-Supervised Learning via Redundancy Reduction

@article{zbontar2021barlow,
  title={Barlow Twins: Self-Supervised Learning via Redundancy Reduction},
  author={Zbontar, Jure and Jing, Li and Misra, Ishan and LeCun, Yann and Deny, St{\'e}phane},
  journal={arXiv preprint arXiv:2103.03230},
  year={2021}
}

Training

This model was trained with the CIFAR training set during 170 epochs . It was then evaluated on the CIFAR validation set by a linear layer trained on top of a frozen Barlow Twins model.

• Training was done on Tesla P100-PCIE-16GB and takes around 1:40 min per epoch

Work is still in progress, here are the current results:

Epochs | Batch Size | Top1 Acc | --- | --- | --- | 170 | 256 | 66.32 |

Original Project Link

This work was inspired by the Facebook Research Github Repository for Barlow Twins