Multiple Object Tracking from appearance by hierarchically clustering tracklets

TL;DR

This paper presents a novel multiple object tracking method that primarily uses appearance features and hierarchical clustering of tracklets, achieving state-of-the-art results on DanceTrack and competitive performance on MOT17 and MOT20.

Contribution

It introduces a hierarchical clustering approach for tracklet fusion based on appearance similarity, emphasizing appearance as the main association cue in MOT.

Findings

Effective on three MOT benchmarks: MOT17, MOT20, and DanceTrack.

Achieves state-of-the-art results on DanceTrack.

Competitive performance on MOT17 and MOT20.

Abstract

Current approaches in Multiple Object Tracking (MOT) rely on the spatio-temporal coherence between detections combined with object appearance to match objects from consecutive frames. In this work, we explore MOT using object appearances as the main source of association between objects in a video, using spatial and temporal priors as weighting factors. We form initial tracklets by leveraging on the idea that instances of an object that are close in time should be similar in appearance, and build the final object tracks by fusing the tracklets in a hierarchical fashion. We conduct extensive experiments that show the effectiveness of our method over three different MOT benchmarks, MOT17, MOT20, and DanceTrack, being competitive in MOT17 and MOT20 and establishing state-of-the-art results in DanceTrack.