Self-supervised Learning of Geometrically Stable Features Through Probabilistic Introspection

TL;DR

This paper introduces a self-supervised method for learning geometrically stable visual features using probabilistic introspection, reducing supervision needed for semantic matching and part detection.

Contribution

It extends self-supervised learning to geometry-oriented tasks by developing a probabilistic approach that identifies stable image regions for dense descriptor learning.

Findings

Requires less supervision for semantic object parts

Pre-trained features excel in semantic matching

Robust probabilistic formulation improves stability

Abstract

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.