Research map / Label-efficient and robust learning / Recognition and long-tail
Noise Modeling, Synthesis and Classification for Generic Object Anti-Spoofing
Joel Stehouwer, Amin Jourabloo, Yaojie Liu, Xiaoming Liu
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 |
|---|---|---|---|
| use GOLab's joint sensor-and-medium classification rather than the strongest reported modality-specific baseline | GOSet anti-spoofing · GOSet test | HTER: 13.4 -> 6.3 | Improved |
| augment GOLab training with GOGen synthetic spoof data (ablation) | GOLab · GOSet test, training data 1/2 live and 1/3 spoof | AUC: 87.6 -> 92.5 | Improved |
| replace learned sensor and medium noise prototypes with constant maps (ablation) | GOGen · GOSet synthetic images | Rank-1 sensor identification accuracy: 56.0% -> 11.0% | Hurt |
| replace learned sensor and medium noise prototypes with constant maps (ablation) | GOGen · GOSet synthetic images | Rank-1 medium identification accuracy: 26.3% -> 19.7% | Hurt |
| use sensor classification loss only instead of both sensor and medium losses (ablation) | GOLab · GOSet test | sensor Rank-1 accuracy: 84.7% -> 87.6% | Hurt |
| use medium classification loss only instead of both sensor and medium losses (ablation) | GOLab · GOSet test | AUC: 85.9 -> 97.5 | Hurt |
| increase training data from the smallest sampled set to all live and spoof data (ablation) | GOLab · GOSet test | AUC: 79.7 -> 97.5 | Improved |
| use GOLab's n-ary sensor/medium classification instead of GOPad binary classification (ablation) | GOSet anti-spoofing · GOSet test, all training data | AUC: 87.6 -> 97.5 | Improved |
| train a GOSet model and evaluate on face images rather than train/evaluate on the target face domain | GOPad · MSU SiW | HTER: 11.0 -> 9.5 | Improved |
| transfer a GOSet-trained model to OULU P1 | GOLab · OULU P1 | HTER: 11.8 -> 41.2 | Hurt |
Closest papers by meaning
- Face De-Spoofing: Anti-Spoofing via Noise Modeling (ECCV 2018 · similarity 0.87)
- Deep Learning for Face Anti-Spoofing: A Survey (TPAMI 2022 · similarity 0.86)
- A Dataset and Benchmark for Large-Scale Multi-Modal Face Anti-Spoofing (CVPR 2019 · similarity 0.85)
- On Disentangling Spoof Trace for Generic Face Anti-Spoofing (ECCV 2020 · similarity 0.85)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.84)
- One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · similarity 0.83)
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.
