Research map / Detection and segmentation / CNN features
UIA-ViT: Unsupervised Inconsistency-Aware Method Based on Vision Transformer for Face Forgery Detection
Wanyi Zhuang, Qi Chu, Zhentao Tan, Qiankun Liu, Haojie Yuan, Changtao Miao, Zixiang Luo, Nenghai Yu
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 UPCL (unsupervised patch consistency learning) (ablation) | ViT-Base · Celeb-DF-v1 | AUC (%): 75.32 -> 82.96 | Improved |
| add UPCL (ablation) | ViT-Base · Celeb-DF-v2 | AUC (%): 76.25 -> 80.86 | Improved |
| replace fixed-GFR PCL supervision with UPCL (MVG-refined pseudo labels) (ablation) | UIA-ViT · Celeb-DF-v1 | AUC (%): 77.88 -> 82.96 | Improved |
| use soft learnable consistency factors instead of fixed ones (UPCL-hard) (ablation) | UIA-ViT · Celeb-DF-v1 | AUC (%): 78.54 -> 82.96 | Improved |
| use soft learnable consistency factors instead of fixed ones (UPCL-hard) (ablation) | UIA-ViT · Celeb-DF-v2 | AUC (%): 78.55 -> 80.86 | Improved |
| add CWA on top of UPCL (ablation) | UIA-ViT · Celeb-DF-v1 | AUC (%): 82.96 -> 84.77 | Improved |
| make assemble progressive (PCWA instead of CWA) (ablation) | UIA-ViT · Celeb-DF-v1 | AUC (%): 84.77 -> 86.59 | Improved |
| make assemble progressive (PCWA instead of CWA) (ablation) | UIA-ViT · Celeb-DF-v2 | AUC (%): 81.77 -> 82.41 | Improved |
| add UPCL and PCWA to ViT-Base (ablation) | ViT-Base · Celeb-DF-v1 | AUC (%): 75.32 -> 86.59 | Improved |
| use patch embeddings of layer 6 rather than layers 5, 7, 8 for MVG estimation (ablation) | UIA-ViT · unseen datasets | AUC (%) | Improved |
| use standard central GFR rather than GFR low/large/small (ablation) | UIA-ViT · unseen datasets | AUC (%) | Improved |
| compare UIA-ViT with prior methods on DFDC-P | UIA-ViT vs DCL · DFDC-P | AUC (%): 75.80 vs 76.71 | Hurt |
Later papers that compared with this method
1 later accepted paper reports UIA-ViT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Real Appearance Modeling for More General Deepfake Detection (ECCV 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
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.88)
- Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face Forgery Detection (CVPR 2025 · similarity 0.86)
- SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints (AAAI 2025 · similarity 0.86)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.86)
- Face Forgery Video Detection via Temporal Forgery Cue Unraveling (CVPR 2025 · similarity 0.86)
- FakeDiffer: Distributional Disparity Learning on Differentiated Reconstruction for Face Forgery Detection (AAAI 2025 · similarity 0.85)
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.
