jack-willturner/self-rep-learning

Self-supervised visual representation learning

Self-Supervised Visual Representation Learning

PyTorch replication of papers studied in this paper.

main.py currently trains a very very small ResNet, and then chops off the last layers and optimises a logistic regression model with L-BFGS to fit the representations generated by our new amputated model.

More models and experiments coming soon...

For now:

• Tensorflow implementation
• PyTorch: Unsupervised Representation Learning by Predicting Image Rotations
• Caffe: Unsupervised Representation Learning By Context Prediction
• PyTorch: Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles
• Caffe: Discriminative Unsupervised Feature Learning with Exemplar Convolutional Neural Networks

Acknowledgements

Code liberally borrowed from:

• https://github.com/jhjacobsen/pytorch-i-revnet