Research map / Label-efficient and robust learning / Recognition and long-tail
Face Anti-Spoofing via Disentangled Representation Learning
Ke-Yue Zhang, Taiping Yao, Jian Zhang, Ying Tai, Shouhong Ding, Jilin Li, Feiyue Huang, Haichuan Song, Lizhuang Ma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use LBP supervision on features and depth on images instead of binary classification (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, liveness features) | ACER (%): 3.64 -> 1.56 | Improved |
| use LBP-plus-depth supervision instead of 0/1 map regression (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, liveness features) | ACER (%): 3.02 -> 1.56 | Improved |
| use different supervisions (LBP and depth) rather than LBP on both (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, liveness features) | ACER (%): 1.87 -> 1.56 | Improved |
| use different supervisions rather than depth on both (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, liveness features) | ACER (%): 1.69 -> 1.56 | Improved |
| order of supervision (LBP on features, depth on images vs reverse) (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, liveness features) | ACER (%): 1.65 -> 1.56 | No clear effect |
| fuse with depth network, LBP-plus-depth vs binary classification (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, score fusion with depth net) | ACER (%): 2.78 -> 1.25 | Improved |
| fuse with depth network, LBP-plus-depth vs Depth-Depth (ablation) | disentanglement framework · Oulu-NPU protocol 1 (ablation, score fusion with depth net) | ACER (%): 1.80 -> 1.25 | Improved |
| average fusion of LBP and depth map scores instead of maximum (ablation) | score fusion · Oulu-NPU protocol 1 | ACER (%): 1.88 -> 1.25 | Improved |
| average fusion instead of LBP map alone (ablation) | score fusion · Oulu-NPU protocol 1 | ACER (%): 1.56 -> 1.25 | Improved |
| average fusion instead of depth map alone (ablation) | score fusion · Oulu-NPU protocol 1 | ACER (%): 1.67 -> 1.25 | Improved |
Closest papers by meaning
- Learning Polysemantic Spoof Trace: A Multi-Modal Disentanglement Network for Face Anti-spoofing (AAAI 2023 · similarity 0.91)
- Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision (CVPR 2018 · similarity 0.90)
- Deep Learning for Face Anti-Spoofing: A Survey (TPAMI 2022 · similarity 0.90)
- SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · similarity 0.88)
- One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · similarity 0.87)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
