Open-set Face Recognition using Ensembles trained on Clustered Data

Rafael Henrique Vareto,William Robson Schwartz

DOI: https://doi.org/10.1109/IJCB48548.2020.9304882

2023-08-15

Abstract:Open-set face recognition describes a scenario where unknown subjects, unseen during the training stage, appear on test time. Not only it requires methods that accurately identify individuals of interest, but also demands approaches that effectively deal with unfamiliar faces. This work details a scalable open-set face identification approach to galleries composed of hundreds and thousands of subjects. It is composed of clustering and an ensemble of binary learning algorithms that estimates when query face samples belong to the face gallery and then retrieves their correct identity. The approach selects the most suitable gallery subjects and uses the ensemble to improve prediction performance. We carry out experiments on well-known LFW and YTF benchmarks. Results show that competitive performance can be achieved even when targeting scalability.

Computer Vision and Pattern Recognition,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

This paper attempts to address the problem of face recognition in open-set scenarios. Specifically, this scenario involves new individuals appearing during the testing phase who were not seen during the training phase. This not only requires the method to accurately recognize the individuals of interest but also to effectively handle these unfamiliar faces.

The paper proposes a scalable open-set face recognition method suitable for face galleries containing hundreds or thousands of individuals. This method combines clustering algorithms and binary classifier ensemble techniques to first estimate whether the query face sample belongs to the face gallery and then retrieve its correct identity. By selecting the most appropriate gallery individuals and using ensemble techniques to improve prediction performance, this method achieves scalability while maintaining accuracy.

### Main Contributions:
1. **Scalable open-set method**: New individuals can be added without retraining.
2. **Easy-to-implement method**: Few parameters, easy to adjust.
3. **Reduced training space**: Improved prediction performance through clustering and learning small-scale ensembles.
4. **Comprehensive experimental discussion**: Different evaluation protocols considered.
5. **Baseline method**: Provides a reference for future research.

### Problems Addressed:
- **Open-set recognition**: How to handle unseen individuals during testing.
- **Scalability**: How to efficiently perform face recognition in large-scale galleries.
- **Accuracy and simplicity trade-off**: How to simplify the model while maintaining high accuracy.

Through these contributions, the paper aims to advance open-set face recognition technology, making it more practical and efficient in real-world applications.