Self-supervised Learning with Summation and Difference of Embedding Vectors

A Self-supervised learning algorithm that uses summation and difference of embedding vectors that are generated from partial and full image.

Algorithm

Loss Function Graph

Blue: Negative Losses Red: Positive Losses

More info: https://tekinevrim.medium.com/self-supervised-learning-algorithm-vector-difference-and-vector-sum-with-ious-vdvs-d8d4ecbdecfb