Towards Grand Unification of Object Tracking

TL;DR

Unicorn is a unified object tracking framework that can simultaneously handle multiple tracking tasks with a single model, outperforming or matching task-specific methods across diverse datasets, marking a step towards a general vision model.

Contribution

The paper introduces Unicorn, the first unified network architecture and learning paradigm capable of solving multiple object tracking tasks simultaneously.

Findings

Performs on par or better than task-specific trackers across 8 datasets.

Uses the same model parameters and architecture for all tasks.

Demonstrates the feasibility of a unified approach to diverse tracking problems.

Abstract

We present a unified method, termed Unicorn, that can simultaneously solve four tracking problems (SOT, MOT, VOS, MOTS) with a single network using the same model parameters. Due to the fragmented definitions of the object tracking problem itself, most existing trackers are developed to address a single or part of tasks and overspecialize on the characteristics of specific tasks. By contrast, Unicorn provides a unified solution, adopting the same input, backbone, embedding, and head across all tracking tasks. For the first time, we accomplish the great unification of the tracking network architecture and learning paradigm. Unicorn performs on-par or better than its task-specific counterparts in 8 tracking datasets, including LaSOT, TrackingNet, MOT17, BDD100K, DAVIS16-17, MOTS20, and BDD100K MOTS. We believe that Unicorn will serve as a solid step towards the general vision model. Code is available at https://github.com/MasterBin-IIAU/Unicorn.