ifzhang
Is there a tutorial for the self-supervised learning method mentioned in the code repository?
Hello @ifzhang Is there a tutorial for the self-supervised learning method mentioned in the code repository?
In the case of re-identification or embedding training, self-supervised learning usually means that you train binary regression model that tries to decide if the inputs on their two branches are the same object or not. This way, if you have a series of images of the same object, you can generate positive pairs (send different images of the same object on the two branches) and negative pairs (send images of different objects to the model) in a semi-automatic way. The prototype of these models are the siamese networks, you can read tons of information about them virtually everywhere.
