CVJul 14, 2025

Glance-MCMT: A General MCMT Framework with Glance Initialization and Progressive Association

arXiv:2507.10115v11 citationsh-index: 3Has Code
Originality Synthesis-oriented
AI Analysis

This addresses tracking consistency in surveillance or multi-view systems, but appears incremental as it builds on existing methods like BoT-SORT.

The paper tackles multi-camera multi-target tracking by proposing a framework that uses glance initialization and progressive association to ensure consistent global identity assignment across views, achieving results with unspecified concrete numbers.

We propose a multi-camera multi-target (MCMT) tracking framework that ensures consistent global identity assignment across views using trajectory and appearance cues. The pipeline starts with BoT-SORT-based single-camera tracking, followed by an initial glance phase to initialize global IDs via trajectory-feature matching. In later frames, new tracklets are matched to existing global identities through a prioritized global matching strategy. New global IDs are only introduced when no sufficiently similar trajectory or feature match is found. 3D positions are estimated using depth maps and calibration for spatial validation.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes