Research map / Label-efficient and robust learning / Recognition and long-tail
PatchNet: A Simple Face Anti-Spoofing Framework via Fine-Grained Patch Recognition
Chien-Yi Wang, Yu-Ding Lu, Shang-Ta Yang, Shang-Hong Lai
Check 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 |
|---|---|---|---|
| reformulate binary face anti-spoofing as fine-grained patch-type recognition using raw-frame patch crops (ablation) | PatchNet · OULU-NPU protocol 1 | ACER (%): 6.25 -> 1.88 | Improved |
| use patch crops instead of resized input for binary classification (ablation) | PatchNet · OULU-NPU protocol 1 | ACER (%): 6.25 -> 3.54 | Improved |
| use fine-grained classes instead of binary classes with resized input (ablation) | PatchNet · OULU-NPU protocol 1 | ACER (%): 6.25 -> 5.63 | Improved |
| add self-supervised similarity loss (ablation) | PatchNet with fine-grained patch recognition · OULU-NPU protocol 1 | ACER (%): 1.88 -> 1.46 | Improved |
| add asymmetric margin classification loss (ablation) | PatchNet with fine-grained patch recognition · OULU-NPU protocol 1 | ACER (%): 1.88 -> 0.63 | Improved |
| add both asymmetric margin classification and self-supervised similarity losses (ablation) | PatchNet · OULU-NPU protocol 1 | ACER (%): 1.88 -> 0.0 | Improved |
| remove angular margin (ablation) | PatchNet · O&C&I to M | HTER (%): 7.10 -> 10.0 | Hurt |
| remove self-supervised similarity loss (ablation) | PatchNet · O&C&I to M | HTER (%): 7.10 -> 8.9 | Hurt |
| replace fine-grained patch classes with coarse classes (ablation) | PatchNet · O&C&I to M | HTER (%): 7.10 -> 10.24 | Hurt |
| use PatchNet rather than the strongest listed baseline | face anti-spoofing · OULU-NPU protocol 1 | ACER (%): 0.2 -> 0.0 | Improved |
| use PatchNet rather than the strongest listed baseline | face anti-spoofing · SiW protocol 3 | ACER (%): 1.52±0.13 -> 2.45±0.45 | Hurt |
| use PatchNet rather than DC-CDN | cross-dataset face anti-spoofing · CASIA-MFSD to Replay-Attack | HTER (%): 6.0 -> 9.9 | Hurt |
Later papers that compared with this method
2 later accepted papers report PatchNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · better on 1 of 1 shared result)
- Rethinking Domain Generalization for Face Anti-Spoofing: Separability and Alignment (CVPR 2023 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Meta Model for Zero- and Few-Shot Face Anti-Spoofing (AAAI 2020 · similarity 0.90)
- One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · similarity 0.90)
- Deep Learning for Face Anti-Spoofing: A Survey (TPAMI 2022 · similarity 0.90)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.89)
- From Intuition to Investigation: A Tool-Augmented Reasoning MLLM Framework for Generalizable Face Anti-Spoofing (CVPR 2026 · similarity 0.89)
- SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · similarity 0.89)
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.
