Research map / Detection and segmentation / Open-set and few-shot detection
A Rich Knowledge Space for Scalable Deepfake Detection
Inho Jung, Hyeongjun Choi, Binh M. Le, Hohyun Na, Simon S. Woo
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 |
|---|---|---|---|
| add CLAM intermediate CLIP features (ablation) | SD2 · Cross-domain | mAUC: 80.24 -> 81.84 | Improved |
| add LD-Face contrastive loss with facial descriptions (ablation) | SD2 · Cross-domain | mAUC: 81.84 -> 84.34 | Improved |
| add LD-Env contrastive loss with environmental descriptions (ablation) | SD2 · Cross-domain | mAUC: 84.34 -> 87.79 | Improved |
| add both dual contrastive losses (LD-Face and LD-Env) (ablation) | SD2 · Cross-domain | mAUC: 5.95% | Improved |
| increase training data from 100K to 3M images (ablation) | SD2 · Cross-domain | mAUC: 82.90 -> 87.79 | Improved |
| increase training data from 100K to 3M images (ablation) | CLIP-LoRA · Cross-domain | mAUC: 78.84 -> 72.12 | Hurt |
| increase training data from 100K to 3M images (ablation) | CLIP-SVD (Effort) · Cross-domain | mAUC: 82.27 -> 81.98 | No clear effect |
| increase training data from 100K to 3M images (ablation) | SPSL · Cross-domain | mAUC: 71.56 -> 75.68 | Improved |
| increase training data from 100K to 3M images (ablation) | CLIP · Cross-domain | mAUC: 69.82 -> 69.83 | No clear effect |
| vary loss weight alpha from 2.0 to 1.5 or 2.5 (ablation) | SD2 · Cross-domain (1M) | mAUC: 84.85 -> 83.45 / 83.02 | Hurt |
| vary LoRA rank from 8 to 4 or 16 (ablation) | SD2 · Cross-domain (1M) | mAUC: 84.85 -> 82.97 / 84.37 | Hurt |
| use SD2 over CLIP-SVD (Effort) on GenImage | SD2 · GenImage | mACC: 79.94 -> 88.50 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AIDE — AIDE: An Automatic Data Engine for Object Detection in Autonomous Driving (CVPR 2024) · better on 1 of 1 shared result
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
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.92)
- Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · similarity 0.90)
- Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · similarity 0.90)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.90)
- Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning (ICLR 2026 · similarity 0.89)
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 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.
