Can Generative Models Improve Self-Supervised Representation Learning?

arXiv:2403.05966v1 Announce Type: cross
Abstract: The rapid advancement in self-supervised learning (SSL) has highlighted its potential to leverage unlabeled data for learning powerful visual representations. However, existing SSL approaches, particularly those employing different views of the same image, often rely on a limited set of predefined data augmentations. This constrains the diversity and quality of transformations, which leads to sub-optimal representations. In this paper, we introduce a novel framework that enriches the SSL paradigm by utilizing generative models to produce …

abstract advancement arxiv cs.cv cs.lg data diversity generative generative models however image representation representation learning self-supervised learning set ssl supervised learning type visual