hanwei0912 / interpretability-methods-for-self-supervised-and-supervised-models Goto Github PK
View Code? Open in Web Editor NEWIn recent years, the rapid development of Deep Neural Networks (DNN) has led to a remarkable performance in many complex tasks in the field of computer vision at the cost of the models’ complexity. The more complex the models get, the higher the need is for understanding them. The primary objective of this repo is to give visual explanations on what both supervised and self-supervised methods really learn during training. Self-supervised and supervised state-of-the-art pre-trained models will be investigated. As backbone networks, for both categories convnets and Transformers based architectures will be used. Variation of visualization techniques will be used.