Deep Constrained Dominant Sets for Person Re-Identification

Explore an in-depth analysis of person re-identification techniques in this 23-minute lecture from the University of Central Florida. Delve into various network architectures, including Triplet Loss, Quadruplet Loss, and Diffusion-based approaches. Learn about the innovative Deep Constrained Dominant Sets (DCDS) method and its implementation in person re-identification tasks. Understand the concept of Dominant Sets Clustering and its constrained variant. Discover the role of Auxiliary Networks and the process of Constraint Expansion in improving re-identification accuracy. Examine the pipeline of DCDS-based networks and evaluate their performance through comprehensive results.

Intro
Overview
Triplet Loss Based Network
Quadruplet Loss Based Network
Diffusion Based Network
DCDS Based Network
Pipeline
Dominant Sets Clustering
Constrained Dominant Sets (CDS)
Auxiliary Net
At Testing
Constraint Expansion
Results