InsCLR: Improving Instance Retrieval with Self-Supervision

TL;DR

InsCLR is a novel self-supervised learning method designed to enhance instance retrieval by dynamically mining pseudo positives to learn invariant representations, outperforming existing SSL methods.

Contribution

This paper introduces InsCLR, a new SSL approach that improves instance retrieval by focusing on intra-class invariance through dynamic pseudo positive mining.

Findings

InsCLR achieves comparable or better performance than state-of-the-art SSL methods.

Dynamic pseudo positive mining enhances robustness to viewpoint and background variations.

InsCLR demonstrates significant improvements in instance retrieval tasks.

Abstract

This work aims at improving instance retrieval with self-supervision. We find that fine-tuning using the recently developed self-supervised (SSL) learning methods, such as SimCLR and MoCo, fails to improve the performance of instance retrieval. In this work, we identify that the learnt representations for instance retrieval should be invariant to large variations in viewpoint and background etc., whereas self-augmented positives applied by the current SSL methods can not provide strong enough signals for learning robust instance-level representations. To overcome this problem, we propose InsCLR, a new SSL method that builds on the \textit{instance-level} contrast, to learn the intra-class invariance by dynamically mining meaningful pseudo positive samples from both mini-batches and a memory bank during training. Extensive experiments demonstrate that InsCLR achieves similar or even better performance than the state-of-the-art SSL methods on instance retrieval. Code is available at https://github.com/zeludeng/insclr.