Cycle-Contrast for Self-Supervised Video Representation Learning

TL;DR

Cycle-Contrastive Learning (CCL) is a self-supervised approach that learns video representations by establishing correspondences across frames and videos, improving performance on downstream video understanding tasks.

Contribution

The paper introduces a novel cycle-contrastive learning method that uniquely models relations between frames and videos within a single network architecture.

Findings

Outperforms previous methods in nearest neighbor retrieval.

Achieves higher accuracy in action recognition tasks.

Demonstrates effective transferability to downstream tasks.

Abstract

We present Cycle-Contrastive Learning (CCL), a novel self-supervised method for learning video representation. Following a nature that there is a belong and inclusion relation of video and its frames, CCL is designed to find correspondences across frames and videos considering the contrastive representation in their domains respectively. It is different from recent approaches that merely learn correspondences across frames or clips. In our method, the frame and video representations are learned from a single network based on an R3D architecture, with a shared non-linear transformation for embedding both frame and video features before the cycle-contrastive loss. We demonstrate that the video representation learned by CCL can be transferred well to downstream tasks of video understanding, outperforming previous methods in nearest neighbour retrieval and action recognition tasks on UCF101, HMDB51 and MMAct.