Research map / Image generation / Autoregressive and latent generation
A Principled Design of Image Representation: Towards Forensic Tasks
Shuren Qi, Yushu Zhang, Chao Wang, Jiantao Zhou, Xiaochun Cao
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 |
|---|---|---|---|
| use DIR instead of FAR/SID descriptors for dense detection | dense detection pipeline · pattern detection | average F1 (%): 67.28 -> 88.20 | Improved |
| use DIR instead of DAISY/DASC for dense matching | dense matching pipeline · INRIA Holidays matching | average repeatability (%): 39.40 -> 52.31 | Improved |
| use DIR under 45 degree rotation instead of DAISY | dense matching pipeline · INRIA Holidays matching | repeatability (%): 0.30 -> 76.61 | Improved |
| use DIR under scaling 0.8 instead of DAISY | dense matching pipeline · INRIA Holidays matching | repeatability (%): 72.07 -> 49.36 | Hurt |
| replace PCT features in Cozzolino'15 with DIR | Cozzolino'15 · UNIPA copy-move | F1 (%): 88.31 -> 95.66 | Improved |
| replace PCT features in Cozzolino'15 with DIR | Cozzolino'15 · FAU copy-move | F1 (%): 91.89 -> 93.62 | Improved |
| replace PCT features in Cozzolino'15 with DIR | Cozzolino'15 · CMH copy-move | F1 (%): 80.12 -> 79.5 | No clear effect |
| replace PCT features with DIR under scaling 0.8 | Cozzolino'15 · FAU copy-scale-move | F1 (%): 56.15 -> 71.18 | Improved |
| replace PCT features with DIR under scaling 0.8 | Cozzolino'15 · FAU copy-scale-move | recall (%): 46.06 -> 63.32 | Improved |
| replace PCT features with DIR under scaling 0.8 | Cozzolino'15 · FAU copy-scale-move | matches per image: 144519.75 -> 270730.23 | Improved |
| replace dense DCT with DIR in SIFT-based perceptual hashing | SIFT-DCT hashing · RTD | average F1 (%): 63.76 -> 69.94 | Improved |
| replace DCT with DIR in hashing, clean images | SIFT-DCT hashing · RTD clean | F1 (%): 79.85 -> 76.95 | Hurt |
Closest papers by meaning
- Forensic Self-Descriptions Are All You Need for Zero-Shot Detection, Open-Set Source Attribution, and Clustering of AI-generated Images (CVPR 2025 · similarity 0.86)
- ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization (NeurIPS 2025 · similarity 0.85)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.84)
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 2025 · similarity 0.84)
- SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints (AAAI 2025 · similarity 0.84)
- On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection (NeurIPS 2024 · similarity 0.84)
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.
