Identity-conditioned fusion improves person re-identification accuracy
Identity-Conditioned Score Fusion for Open-Set Person Re-Identification
Computer Vision and Pattern RecognitionMachine Learning
Summary
People can be recognized across different camera views by combining clues like face, walking style, and body shape. The authors found that adapting how these clues are combined for each person in a database, without any extra training, helps tell true matches from impostors better. This method balances the matching scores well and works especially well even when people change clothes. Tests on several challenging datasets showed their approach reduces mistakes in identifying the same person by up to 8.8%.
What this means in practice
- •For security system developers: Enhance multi-camera tracking by dynamically adjusting recognition scores per individual to improve matching accuracy, especially when clothing changes.$Commercial implications: Enables building more reliable person re-identification products for surveillance and access control systems that adapt to individual recognition patterns.
- •For robotics engineers: Improve human identification in robots by tailoring sensor data fusion per person without retraining models, enhancing identification in real-world environments.
Authors
Manyi Yao, Jurijs Nazarovs, Eunji Chong, Abhishek Sharma, Rohan Sarkar, Yue Guo, Christian R. Shelton, Amit K. Roy-Chowdhury, Debashish Pal
Abstract
Robust person re-identification often combines complementary cues such as face, gait, and body shape. While adaptive fusion typically targets query quality, model strength also varies across identities. We introduce identity-conditioned score fusion, a framework that tailors weights to each gallery identity without training. By contrasting intra-identity consistency against cross-identity impostors, it extracts identity-specific profiles that couple with query-conditioned adaptation via a parameter-free rule. This widens the separation between true and false matches while preserving score calibration. Evaluations on three clothes-changing person re-identification benchmarks show that our method consistently outperforms statistical, rank-based, and learned baselines, achieving up to an 8.8% absolute reduction in the false non-identification rate and demonstrating the value of identity-conditioned fusion in open-set person re-identification.