Video Representation Learning with Visual Tempo Consistency

TL;DR

This paper introduces a self-supervised video representation learning method using visual tempo consistency, leveraging hierarchical contrastive learning to improve action recognition and other downstream tasks.

Contribution

It proposes a novel hierarchical contrastive learning approach that uses visual tempo as a self-supervision signal for video representation learning.

Findings

Achieves 82.1% on UCF-101 action recognition

Generalizes well to action detection and anticipation tasks

Introduces Instance Correspondence Map for visualization

Abstract

Visual tempo, which describes how fast an action goes, has shown its potential in supervised action recognition. In this work, we demonstrate that visual tempo can also serve as a self-supervision signal for video representation learning. We propose to maximize the mutual information between representations of slow and fast videos via hierarchical contrastive learning (VTHCL). Specifically, by sampling the same instance at slow and fast frame rates respectively, we can obtain slow and fast video frames which share the same semantics but contain different visual tempos. Video representations learned from VTHCL achieve the competitive performances under the self-supervision evaluation protocol for action recognition on UCF-101 (82.1\%) and HMDB-51 (49.2\%). Moreover, comprehensive experiments suggest that the learned representations are generalized well to other downstream tasks including action detection on AVA and action anticipation on Epic-Kitchen. Finally, we propose Instance Correspondence Map (ICM) to visualize the shared semantics captured by contrastive learning.