A simple PyTorch implementation of the SimSiam algorithm from Exploring Simple Siamese Representation Learning. In addition to the implementation of the SimSiam algorithm, this repository contains a simple experiment which compares the training results from SimSiam with and without the stop-gradient operation in the algorithm. This comparison was also performed in Exploring Simple Siamese Representation Learning.
The dataset used to generate the results in the report was the tiny-imagenet-200 dataset from ImageNet.
- PyTorch
- Matplotlib
- Numpy