Although deep learning is a powerful tool that has progressed in the field of computer vision, its success is attributed to supervised learning tasks. Supervised learning tasks require manual annotations which are time-consuming and expensive to acquire.

In addition, models trained on supervised learning are unable to generalize well to unseen examples. In recent years, self-supervised learning techniques have been used to address these problems and have achieved several state-of-the-art (SOTA) performances on several tasks.