Self-Supervised Visual Representations Learning by Contrastive Mask Prediction

TL;DR

This paper introduces a contrastive mask prediction method for self-supervised visual learning, which contrasts region features instead of view features, improving performance on diverse datasets and downstream tasks.

Contribution

It proposes a novel contrastive mask prediction task and a mask contrast framework that relaxes semantic assumptions and enhances generalization in self-supervised learning.

Findings

MaskCo achieves comparable results to MoCo V2 on ImageNet.

MaskCo outperforms MoCo V2 on COCO and Conceptual Captions datasets.

Region-level contrastive learning improves downstream task performance.

Abstract

Advanced self-supervised visual representation learning methods rely on the instance discrimination (ID) pretext task. We point out that the ID task has an implicit semantic consistency (SC) assumption, which may not hold in unconstrained datasets. In this paper, we propose a novel contrastive mask prediction (CMP) task for visual representation learning and design a mask contrast (MaskCo) framework to implement the idea. MaskCo contrasts region-level features instead of view-level features, which makes it possible to identify the positive sample without any assumptions. To solve the domain gap between masked and unmasked features, we design a dedicated mask prediction head in MaskCo. This module is shown to be the key to the success of the CMP. We evaluated MaskCo on training datasets beyond ImageNet and compare its performance with MoCo V2. Results show that MaskCo achieves comparable performance with MoCo V2 using ImageNet training dataset, but demonstrates a stronger performance across a range of downstream tasks when COCO or Conceptual Captions are used for training. MaskCo provides a promising alternative to the ID-based methods for self-supervised learning in the wild.