Self-supervised Co-training for Video Representation Learning

TL;DR

This paper introduces a self-supervised co-training method for video representation learning that leverages multiple views to enhance contrastive learning, achieving state-of-the-art results efficiently on action recognition and video retrieval tasks.

Contribution

It proposes a novel co-training scheme that exploits complementary views like RGB and optical flow to improve contrastive learning in video representations.

Findings

Achieves state-of-the-art or comparable performance on downstream tasks.

Requires less training data for similar performance.

Enhances contrastive learning with multi-view co-training.

Abstract

The objective of this paper is visual-only self-supervised video representation learning. We make the following contributions: (i) we investigate the benefit of adding semantic-class positives to instance-based Info Noise Contrastive Estimation (InfoNCE) training, showing that this form of supervised contrastive learning leads to a clear improvement in performance; (ii) we propose a novel self-supervised co-training scheme to improve the popular infoNCE loss, exploiting the complementary information from different views, RGB streams and optical flow, of the same data source by using one view to obtain positive class samples for the other; (iii) we thoroughly evaluate the quality of the learnt representation on two different downstream tasks: action recognition and video retrieval. In both cases, the proposed approach demonstrates state-of-the-art or comparable performance with other self-supervised approaches, whilst being significantly more efficient to train, i.e. requiring far less training data to achieve similar performance.