Self-Supervised Any-Point Tracking by Contrastive Random Walks

TL;DR

This paper introduces a self-supervised method for point tracking in videos using contrastive random walks and transformers, achieving high accuracy without extensive annotations.

Contribution

It proposes a novel self-supervised approach leveraging contrastive random walks and transformer-based global matching for point tracking.

Findings

Outperforms previous self-supervised methods like DIFT.

Achieves competitive results with supervised tracking methods.

Uses cycle consistency and data augmentation for training stability.

Abstract

We present a simple, self-supervised approach to the Tracking Any Point (TAP) problem. We train a global matching transformer to find cycle consistent tracks through video via contrastive random walks, using the transformer's attention-based global matching to define the transition matrices for a random walk on a space-time graph. The ability to perform "all pairs" comparisons between points allows the model to obtain high spatial precision and to obtain a strong contrastive learning signal, while avoiding many of the complexities of recent approaches (such as coarse-to-fine matching). To do this, we propose a number of design decisions that allow global matching architectures to be trained through self-supervision using cycle consistency. For example, we identify that transformer-based methods are sensitive to shortcut solutions, and propose a data augmentation scheme to address them. Our method achieves strong performance on the TapVid benchmarks, outperforming previous self-supervised tracking methods, such as DIFT, and is competitive with several supervised methods.