Person Re-Identification and Tracking in Multiple Non-Overlapping Cameras - Keynote

Explore person re-identification and tracking across multiple non-overlapping cameras in this 52-minute keynote presentation from the University of Central Florida. Delve into challenges, classification networks, and advanced techniques like SPREID and human semantic parsing. Learn about triplet and quadruplet loss-based networks, diffusion-based approaches, and DCDS networks. Examine the pipeline for person re-identification, including dominant sets clustering, constrained dominant sets, and auxiliary networks. Discover various camera configurations for tracking, including fixed, overlapping, and moving cameras. Gain insights into cross-camera track association, constraint expansion, and track refinement techniques. Conclude with a summary of experimental results for multiple camera setups in this comprehensive overview of cutting-edge person re-identification and tracking methodologies.