GenTrack: A New Generation of Multi-Object Tracking

TL;DR

GenTrack addresses the challenge of robust multi-object tracking with a time-varying number of targets by fusing stochastic PSO-guided particle refinement with deterministic data association. It introduces a hybrid MOT framework that uses PSO to steer particles toward distribution modes, while deterministic birth/death and Hungarian-based data association manage track identities and target initialization. The approach leverages social interactions among targets to enhance particle updates and reduce ID switches, culminating in three variants—GenTrack Basic, GenTrack PSO, and GenTrack PSO-Social—that share a unified state/observation model incorporating space, appearance, detection confidence, penalties, and social scores. Experimental results on MOT17-04 (human) and MooTrack360 (cow) demonstrate superior performance and real-time viability, with PSO-Social achieving strong ID-maintenance and low latency, suggesting practical applicability to real-world tracking scenarios. The work provides a public, minimal-dependency codebase enabling flexible reimplementation and extension to broader MOT tasks, including potential 3D and multi-camera extensions.”

Abstract

This paper introduces a novel multi-object tracking (MOT) method, dubbed GenTrack, whose main contributions include: a hybrid tracking approach employing both stochastic and deterministic manners to robustly handle unknown and time-varying numbers of targets, particularly in maintaining target identity (ID) consistency and managing nonlinear dynamics, leveraging particle swarm optimization (PSO) with some proposed fitness measures to guide stochastic particles toward their target distribution modes, enabling effective tracking even with weak and noisy object detectors, integration of social interactions among targets to enhance PSO-guided particles as well as improve continuous updates of both strong (matched) and weak (unmatched) tracks, thereby reducing ID switches and track loss, especially during occlusions, a GenTrack-based redefined visual MOT baseline incorporating a comprehensive state and observation model based on space consistency, appearance, detection confidence, track penalties, and social scores for systematic and efficient target updates, and the first-ever publicly available source-code reference implementation with minimal dependencies, featuring three variants, including GenTrack Basic, PSO, and PSO-Social, facilitating flexible reimplementation. Experimental results have shown that GenTrack provides superior performance on standard benchmarks and real-world scenarios compared to state-of-the-art trackers, with integrated implementations of baselines for fair comparison. Potential directions for future work are also discussed. The source-code reference implementations of both the proposed method and compared-trackers are provided on GitHub: https://github.com/SDU-VelKoTek/GenTrack

Figures (4)

• Figure 1: An illustration of visual human tracking, with particle visualization.
• Figure 2: ATA-IDF1-HOTA comparisons of trackers on human tracking.
• Figure 3: An illustration of visual cow tracking, without particle visualization.
• Figure 4: ATA-IDF1-HOTA comparisons of trackers on cow tracking.