Variational Self-Supervised Contrastive Learning Using Beta Divergence For Face Understanding

arXiv:2312.00824v2 Announce Type: replace-cross
Abstract: Learning a discriminative semantic space using unlabelled and noisy data remains unaddressed in a multi-label setting. We present a contrastive self-supervised learning method which is robust to data noise, grounded in the domain of variational methods. The method (VCL) utilizes variational contrastive learning with beta-divergence to learn robustly from unlabelled datasets, including uncurated and noisy datasets. We demonstrate the effectiveness of the proposed method through rigorous experiments including linear evaluation and fine-tuning scenarios with multi-label …

abstract arxiv beta cs.cv cs.lg data divergence domain face noise robust self-supervised learning semantic space supervised learning type understanding