Related Experiment Video
Updated: Aug 10, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Open Set face recognition using transduction
1Department of Computer Science, George Mason University, Fairfax, VA 22030, USA. fli@cs.gmu.edu
IEEE Transactions on Pattern Analysis and Machine Intelligence
|November 16, 2005
Summary
This study introduces Open Set TCM-kNN for face recognition, enabling systems to identify known individuals or reject unknown ones. This novel approach enhances security by handling open-set scenarios effectively.
Area of Science:
- Computer Science
- Biometrics
- Machine Learning
Background:
- Open Set face recognition challenges systems to identify known individuals while rejecting unknown ones.
- Existing methods struggle with scenarios where test probes may not have corresponding gallery entries.
Purpose of the Study:
- To introduce a novel transductive inference method for Open Set face recognition.
- To present Open Set TCM-kNN (Transduction Confidence Machine-k Nearest Neighbors) for multiclass authentication with rejection capabilities.
Main Methods:
- Utilizing transductive inference and Kolmogorov complexity for likelihood ratio estimation.
- Developing Open Set TCM-kNN for local estimation in detection tasks.
- Applying Pattern Specific Error Inhomogeneities (PSEI) analysis for error structure investigation.
Main Results:
- Demonstrated feasibility, robustness, and advantages of Open Set TCM-kNN on Open Set identification and watch list tasks using FERET data.
- Showcased suitability of Open Set TCM-kNN for PSEI analysis to identify difficult-to-recognize faces.
- Confirmed PSEI analysis improves biometric performance by addressing error-causing face patterns.
Conclusions:
- Open Set TCM-kNN effectively addresses the Open Set face recognition task, including rejection of unknown identities.
- The method provides a robust solution for multiclass authentication scenarios with an "none of the above" option.
- PSEI analysis integrated with Open Set TCM-kNN enhances biometric system performance by targeting specific error patterns.
