Day 187(Face Recognition) -- SimSiam code understanding(Level2)

In the earlier post, we have seen how data augmentation can be done with various transformations to output two variations for the same image. For this blog, we can see how the backbone, projection and prediction layers are added for model training. Followed by how can we measure the accuracy of the model during the evaluation stage. Model definition & Training: The backbone used here is'resnet50'. Next to the projection head is the prediction layer that has Linear, BatchNorm, ReLU and then a Linear layer.