- tensorflow 2.x
pretrained_weights : pretrained weights of simclr model.
images : contains images like plots.
simclr-data-creator.ipynb : file contains code for creating a custom imagenet data.
simclr-model.ipynb : file contains code for training simclr model.
simclr-linear_evaluator-lg.ipynb : file contains code for the linear evaluation of the trained simclr model, with logistic regression.
simclr-nonlinear-evaluator.ipynb : file contains code for nonlinear evaluation of trained simclr moel, with None linear dense tensorflow layer.
resnet-supervised-model.ipynb : file contains code for creating a supervised model for the custom imagenet data.
- Loss function comes from : https://github.com/sthalles/SimCLR-tensorflow/blob/master/utils/losses.py.
- Data Creation function comes from: https://github.com/mf1024/ImageNet-Datasets-Downloader
Dataset, is created using the simclr-data-creator.ipynb file, which creates a custom imagenet data.
- https://arxiv.org/abs/2002.05709
- https://www.youtube.com/watch?v=lW0ZIOyhRic
- https://www.v7labs.com/blog/self-supervised-learning-guide
- https://amitness.com/2020/03/illustrated-simclr/
- https://sthalles.github.io/simple-self-supervised-learning/
- https://github.com/sthalles/SimCLR
- https://github.com/sayakpaul/SimCLR-in-TensorFlow-2
- https://github.com/garder14/simclr-tensorflow2