Self-supervised Learning of Geometrically Stable Features Through Probabilistic Introspection
This addresses the need for reduced manual labeling in geometry-oriented computer vision tasks, though it builds incrementally on existing unsupervised landmark detection ideas.
The paper tackled the problem of extending self-supervised learning to geometry-oriented tasks like semantic matching and part detection by learning dense distinctive visual descriptors from unlabelled images using synthetic transformations and a robust probabilistic formulation. The result showed that a network pre-trained this way requires significantly less supervision for learning semantic object parts and performs excellently for semantic object matching.
Self-supervision can dramatically cut back the amount of manually-labelled data required to train deep neural networks. While self-supervision has usually been considered for tasks such as image classification, in this paper we aim at extending it to geometry-oriented tasks such as semantic matching and part detection. We do so by building on several recent ideas in unsupervised landmark detection. Our approach learns dense distinctive visual descriptors from an unlabelled dataset of images using synthetic image transformations. It does so by means of a robust probabilistic formulation that can introspectively determine which image regions are likely to result in stable image matching. We show empirically that a network pre-trained in this manner requires significantly less supervision to learn semantic object parts compared to numerous pre-training alternatives. We also show that the pre-trained representation is excellent for semantic object matching.