This project was made for course DD2412 Advanced Deep Learning KTH.
In this project we aim to reproduce the SimCLR framework made by google, but with some limitations due time constrains.
Extension to the paper was done which is hown in the following list:
- Trying a new augmentation strategy RandAugment.
- Experiment with an alternative Lifted structure loss function.
- Comparison with a new state-of-the-art model BYOL.
Project members: