Research map / Vision-language and multimodal / Cross-modal learning
Learning Polysemantic Spoof Trace: A Multi-Modal Disentanglement Network for Face Anti-spoofing
Kaicheng Li, Hongyu Yang, Binghui Chen, Pengyu Li, Biao Wang, Di Huang
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 RGB-D polysemantic spoof trace disentanglement with cross-modal feature fusion | Face anti-spoofing · WMCA GrandTest | ACER: 0.52 -> 0.27 | Improved |
| use RGB-D polysemantic spoof trace disentanglement with cross-modal feature fusion | Face anti-spoofing · WMCA LOO | Mean ACER (%): 7.1±9.9 -> 4.8±6.6 | Improved |
| use RGB-D polysemantic spoof trace disentanglement with cross-modal feature fusion | Face anti-spoofing · CeFA LOO Protocol 1 | ACER (%): 6.8±3.5 -> 5.0±1.9 | Improved |
| use RGB-D polysemantic spoof trace disentanglement with cross-modal feature fusion | Face anti-spoofing · CeFA LOO Protocol 2 | ACER (%): 1.2±0.6 -> 0.9±0.2 | Improved |
| use RGB-D polysemantic spoof trace disentanglement with cross-modal feature fusion | Face anti-spoofing · CeFA LOO Protocol 4 | ACER (%): 9.7±6.1 -> 6.2±1.0 | Improved |
| add spoof-trace disentanglement to the original RGB-D classifier (ablation) | Face anti-spoofing · WMCA GrandTest | ACER: 9.0 -> 5.7 | Improved |
| add the depth modality alongside the RGB modality (ablation) | RGB Only disentanglement · WMCA GrandTest | ACER: 5.7 -> 0.27 | Improved |
| add the RGB modality alongside the depth modality (ablation) | DEP Only disentanglement · WMCA GrandTest | ACER: 3.8 -> 0.27 | Improved |
| replace vanilla feature concatenation with the proposed cross-modal fusion module (ablation) | RGB-D disentanglement · WMCA GrandTest | ACER: 1.1 -> 0.27 | Improved |
| replace original SA-Gate with the proposed cross-modal fusion module (ablation) | RGB-D disentanglement · WMCA GrandTest | ACER: 0.52 -> 0.27 | Improved |
| add center loss (ablation) | RGB-D spoof trace disentanglement · WMCA LOO | Mean ACER (%): 12.4±13.6 -> 4.8±6.6 | Improved |
| add identity consistency loss (ablation) | RGB-D spoof trace disentanglement · WMCA LOO | Mean ACER (%): 7.9±9.3 -> 4.8±6.6 | Improved |
Closest papers by meaning
- On Disentangling Spoof Trace for Generic Face Anti-Spoofing (ECCV 2020 · similarity 0.92)
- Face Anti-Spoofing via Disentangled Representation Learning (ECCV 2020 · similarity 0.91)
- Spoof Trace Disentanglement for Generic Face Anti-Spoofing (TPAMI 2022 · similarity 0.90)
- PatchNet: A Simple Face Anti-Spoofing Framework via Fine-Grained Patch Recognition (CVPR 2022 · similarity 0.89)
- DiffFAS: Face Anti-Spoofing via Generative Diffusion Models (ECCV 2024 · similarity 0.88)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.86)
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.
