Conference proceeding
Open-Set Biometrics: Beyond Good Closed-Set Models
Computer Vision – ECCV 2024, v 15120, pp 243-261
31 Oct 2024
Abstract
Biometric recognition has primarily addressed closed-set identification, assuming all probe subjects are in the gallery. However, most practical applications involve open-set biometrics, where probe subjects may or may not be present in the gallery. This poses distinct challenges in effectively distinguishing individuals in the gallery while minimizing false detections. While it is commonly believed that powerful biometric models can excel in both closed- and open-set scenarios, existing loss functions are inconsistent with open-set evaluation. They treat genuine (mated) and imposter (non-mated) similarity scores symmetrically and neglect the relative magnitudes of imposter scores. To address these issues, we simulate open-set evaluation using minibatches during training and introduce novel loss functions: (1) the identification-detection loss optimized for open-set performance under selective thresholds and (2) relative threshold minimization to reduce the maximum negative score for each probe. Across diverse biometric tasks, including face recognition, gait recognition, and person re-identification, our experiments demonstrate the effectiveness of the proposed loss functions, significantly enhancing open-set performance while positively impacting closed-set performance. Our code and models are available here.
Metrics
10 Record Views
Details
- Title
- Open-Set Biometrics: Beyond Good Closed-Set Models
- Creators
- Yiyang Su - Michigan State UniversityMinchul Kim - Michigan State UniversityFeng Liu - Drexel University, Computer ScienceAnil Jain - Michigan State UniversityXiaoming Liu - Michigan State University
- Publication Details
- Computer Vision – ECCV 2024, v 15120, pp 243-261
- Series
- Lecture Notes in Computer Science
- Publisher
- Springer Nature; Cham
- Number of pages
- 19
- Resource Type
- Conference proceeding
- Language
- English
- Academic Unit
- Computer Science
- Web of Science ID
- WOS:001353685900014
- Scopus ID
- 2-s2.0-85208576053
- Other Identifier
- 991022008296404721
InCites Highlights
Data related to this publication, from InCites Benchmarking & Analytics tool:
- Web of Science research areas
- Computer Science, Artificial Intelligence
- Computer Science, Interdisciplinary Applications
- Computer Science, Theory & Methods