Research map / Image generation / Autoregressive and latent generation
Proactive Image Manipulation Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a learned template to real images and detect manipulation by recovering template similarity | image manipulation detection · 12 unseen GMs | average precision (%): 82.97 -> 92.43 | Helped |
| Train with both STGAN and AutoGAN data instead of STGAN alone | proactive template detector · CycleGAN | average precision (%): 96.12 -> 100 | Helped |
| Train with both STGAN and AutoGAN data instead of STGAN alone | proactive template detector · GauGAN | average precision (%): 91.62 -> 99.69 | Helped |
| Increase template set size (ablation) | proactive template detector · image manipulation detection | average precision | Helped |
| Increase template strength (ablation) | proactive template detector · CycleGAN, StarGAN, and GauGAN | average precision | Helped |
| Remove the pair-wise set distribution loss (ablation) | proactive template detector · CycleGAN | average precision (%): 96.12 -> 66.60 | Hurt |
| Remove the recovery loss (ablation) | proactive template detector · CycleGAN | average precision (%): 96.12 -> 51.59 | Hurt |
| Use a fixed rather than learned template set (ablation) | proactive template detector · CycleGAN | average precision (%): 96.12 -> 46.93 | Hurt |
| Replace the template-recovery encoder with direct classification (ablation) | proactive template detector · CycleGAN | average precision (%): 96.12 -> 50.00 | Hurt |
| Remove the magnitude loss (ablation) | proactive template detector · GauGAN | average precision (%): 91.62 -> 87.44 | Hurt |
| Remove the content-independent template loss (ablation) | proactive template detector · GauGAN | average precision (%): 91.62 -> 80.54 | Hurt |
| Remove the separation loss (ablation) | proactive template detector · GauGAN | average precision (%): 91.62 -> 64.06 | Hurt |
Closest papers by meaning
- GIM: A Million-scale Benchmark for Generative Image Manipulation Detection and Localization (AAAI 2025 · similarity 0.86)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.84)
- IMDPrompter: Adapting SAM to Image Manipulation Detection by Cross-View Automated Prompt Learning (ICLR 2025 · similarity 0.84)
- A New Benchmark and Model for Challenging Image Manipulation Detection (AAAI 2024 · similarity 0.84)
- IDseq: Decoupled and Sequentially Detecting and Grounding Multi-Modal Media Manipulation (AAAI 2025 · similarity 0.83)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · 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.