Taming Self-Supervised Learning for Presentation Attack Detection: De-Folding and De-Mixing

Keyphrases

• Baseline Performance 25%
• Biometric System 25%
• CASIA-FASD 25%
• Challenging Problems 25%
• Comprehensive Representation 25%
• Deep Learning 25%
• Demixing 100%
• Detection Method 25%
• Detection Model 25%
• Equal Error Rate 25%
• Global Information 25%
• Global-local 25%
• Handcrafted Features 25%
• Hybrid Data 25%
• Instance-based 25%
• Interpolation Method 25%
• Learning-based 25%
• Local Patterns 25%
• Local Perspectives 25%
• Presentation Attack 50%
• Presentation Attack Detection 100%
• PSA Detection 100%
• Replay Attack 25%
• Self-supervised Learning 100%
• Source Code 25%
• Specific Features 50%
• State-of-the-art Techniques 25%
• Task-specific Representation 25%

Computer Science

• Biometric System 50%
• Deep Learning 50%
• Equal Error Rate 50%
• Experimental Result 50%
• Global Information 50%
• Presentation Attack Detection 100%
• Self-Supervised Learning 100%
• Source Codes 50%