Learning Pixel Trajectories with Multiscale Contrastive Random Walks

TL;DR

This paper introduces a multiscale contrastive random walk approach for pixel-level space-time correspondence, unifying various video modeling tasks like optical flow and object tracking under a self-supervised framework.

Contribution

It extends contrastive random walks to dense pixel graphs with a hierarchical, coarse-to-fine search, enabling a unified self-supervised learning method for multiple video tasks.

Findings

Achieves competitive performance on optical flow, keypoint tracking, and segmentation.

Unifies multiple tasks with a single self-supervised model.

Demonstrates effectiveness of multiscale hierarchy in space-time correspondence.

Abstract

A range of video modeling tasks, from optical flow to multiple object tracking, share the same fundamental challenge: establishing space-time correspondence. Yet, approaches that dominate each space differ. We take a step towards bridging this gap by extending the recent contrastive random walk formulation to much denser, pixel-level space-time graphs. The main contribution is introducing hierarchy into the search problem by computing the transition matrix between two frames in a coarse-to-fine manner, forming a multiscale contrastive random walk when extended in time. This establishes a unified technique for self-supervised learning of optical flow, keypoint tracking, and video object segmentation. Experiments demonstrate that, for each of these tasks, the unified model achieves performance competitive with strong self-supervised approaches specific to that task. Project webpage: https://jasonbian97.github.io/flowwalk