Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 11·5 min read
A Combined Feature-Based Framework for Disguise and Spoofing Detection in Face Recognition Systems
TL;DR
A new framework for face recognition systems effectively detects both disguise and spoofing attempts by combining various feature-extraction methods. The Histogram of Oriented Gradients (HOG) pipeline showed the best overall performance, achieving 94.59% accuracy in mixed-appearance scenarios.
✦ Why It Matters
Consider integrating the HOG-based pipeline for improved face recognition accuracy in your applications.
Key Takeaways
- Explore the HOG-based pipeline for better disguise detection.
- Evaluate the LBP-based pipeline for high spoofing accuracy.
- Test face recognition systems under varied conditions for robustness.
- Consider deep learning methods for future enhancements.
How It Works
The proposed framework utilizes a two-phase process involving pre-processing, feature extraction, filtering, and classification. Each pipeline employs different feature-extraction techniques, such as PCA, Local Binary Patterns, and Histogram of Oriented Gradients, to effectively differentiate between legitimate users and spoofing attempts.