Tracking Emerges by Colorizing Videos

TL;DR

This paper introduces a self-supervised approach to visual tracking by training a model to colorize grayscale videos, which naturally encourages the model to learn tracking without manual labels, outperforming some existing methods.

Contribution

The paper presents a novel self-supervised method that leverages video colorization to learn visual tracking without labeled data, outperforming optical flow-based methods.

Findings

The model learns to track visual regions effectively through colorization.

It outperforms recent optical flow-based tracking methods.

Failures in tracking correlate with failures in colorization.

Abstract

We use large amounts of unlabeled video to learn models for visual tracking without manual human supervision. We leverage the natural temporal coherency of color to create a model that learns to colorize gray-scale videos by copying colors from a reference frame. Quantitative and qualitative experiments suggest that this task causes the model to automatically learn to track visual regions. Although the model is trained without any ground-truth labels, our method learns to track well enough to outperform the latest methods based on optical flow. Moreover, our results suggest that failures to track are correlated with failures to colorize, indicating that advancing video colorization may further improve self-supervised visual tracking.