Causal Bootstrapped Alignment for Unsupervised Video-Based Visible–Infrared Person Re-Identification
Summary
This research addresses video-based visible-infrared person re-identification (VVI-ReID, a technique that identifies the same person across visible light and thermal infrared video footage for surveillance) using unsupervised learning, which doesn't require expensive labeled training data. The authors propose Causal Bootstrapped Alignment (CBA), a framework that uses temporal video information and causal intervention (a method that identifies cause-and-effect relationships by simulating changes) to improve how well the system recognizes people across both imaging modes.
Solution / Mitigation
The source presents the proposed CBA framework as the solution, which includes two components: (1) Causal Intervention Warm-up (CIW) that 'performs sequence-level causal interventions by leveraging temporal identity consistency and cross-modality identity consistency to suppress modality- and motion-induced spurious correlations while preserving identity-relevant semantics,' and (2) Prototype-Guided Uncertainty Refinement (PGUR) that 'employs a coarse-to-fine alignment strategy to resolve cross-modality granularity mismatch.' Code is available at https://github.com/Visuang/CBA.
Classification
Original source: http://ieeexplore.ieee.org/document/11579407
First tracked: July 16, 2026 at 02:12 AM
Classified by LLM (prompt v3) · confidence: 95%