Video Representation Learning with Visual Tempo Consistency
This work addresses video representation learning for action recognition and other downstream tasks, offering a novel self-supervision approach that is incremental in leveraging visual tempo.
The paper tackles video representation learning by using visual tempo as a self-supervision signal, proposing hierarchical contrastive learning (VTHCL) to maximize mutual information between slow and fast video representations, achieving competitive results like 82.1% on UCF-101 and 49.2% on HMDB-51 for action recognition.
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.