Research map / Detection and segmentation / Open-set and few-shot detection
Noise-assisted Prompt Learning for Image Forgery Detection and Localization
Dong Li, Jiaying Zhu, Xueyang Fu, Xun Guo, Yidi Liu, Gang Yang, Jiawei Liu, Zheng-Jun Zha
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 prompt adjustment network (PANet) (ablation) | CLIP-IFDL · CASIA / NIST16 (fine-tuned) | AUC: 83.3 -> 91.3 (CASIA); 84.4 -> 99.8 (NIST16) | Improved |
| add embedding adjustment network (EANet) (ablation) | CLIP-IFDL · CASIA / NIST16 (fine-tuned) | AUC: 84.2 -> 91.3 (CASIA); 88.9 -> 99.8 (NIST16) | Improved |
| add forgery-enhanced noise adapter (FENA) (ablation) | CLIP-IFDL · CASIA / NIST16 (fine-tuned) | AUC: 88.6 -> 91.3 (CASIA); 94.9 -> 99.8 (NIST16) | Improved |
| unfreeze all CLIP parameters (ablation) | CLIP-IFDL · IMD20 (pre-trained) | AUC: 97.8 -> 75.4 | Hurt |
| unfreeze CLIP image or text encoder only (ablation) | CLIP-IFDL · IMD20 (pre-trained) | AUC: 97.8 -> 86.7 (image), 90.1 (text) | Hurt |
| use CLIP-IFDL instead of HiFi-Net in pre-trained setting | Columbia · Columbia (pre-trained) | AUC: 98.3 -> 97.6 | Hurt |
| use CLIP-IFDL under Gaussian noise and other distortions | NIST16 localization · NIST16 (distorted) | AUC: 89.68 -> 86.75 at noise sigma=15, versus ObjectFormer 78.15 | Improved |
| use CLIP-IFDL for detection on challenging COVER and IMD20 | detection · COVER / IMD20 | F1: 65.2 vs 49.8 (COVER); 71.2 vs 52.1 (IMD20) | 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.
- SAFL-Net — SAFL-Net: Semantic-Agnostic Feature Learning Network with Auxiliary Plugins for Image Manipulation Detection (ICCV 2023) · better on 10 of 10 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 Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.88)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.88)
- Reinforced Multi-teacher Knowledge Distillation for Efficient General Image Forgery Detection and Localization (AAAI 2025 · similarity 0.88)
- Forgery-aware Adaptive Transformer for Generalizable Synthetic Image Detection (CVPR 2024 · similarity 0.88)
- A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization (AAAI 2025 · similarity 0.87)
- Unlocking the Capabilities of Large Vision-Language Models for Generalizable and Explainable Deepfake Detection (ICML 2025 · similarity 0.87)
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.
