Research map / Detection and segmentation / CNN features
Learning Rich Features for Image Manipulation Detection
Read the paperCheck 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 |
|---|---|---|---|
| combine RGB and SRM-noise features with bilinear pooling (ablation) | Faster R-CNN manipulation detector · NIST16, COVER, CASIA | F1 score: 0.567 -> 0.722; 0.391 -> 0.437; 0.392 -> 0.408 | Helped |
| combine RGB and SRM-noise features with bilinear pooling (ablation) | Faster R-CNN manipulation detector · Columbia | F1 score: 0.585 -> 0.697; 0.705 -> 0.697 | Mixed |
| replace late fusion of detected boxes with bilinear feature pooling (ablation) | RGB-N · NIST16, Columbia, COVER, CASIA | F1 score: 0.625 -> 0.722; 0.681 -> 0.697; 0.371 -> 0.437; 0.397 -> 0.408 | Helped |
| use RGB features for RPN proposals instead of noise features (ablation) | two-stream Faster R-CNN · Synthetic COCO test | AP: 0.472 -> 0.627 | Helped |
| use both RGB and noise features as RPN input rather than RGB features only (ablation) | two-stream Faster R-CNN · Synthetic COCO test | AP: 0.620 -> 0.627 | No clear effect |
| add image flipping augmentation (ablation) | RGB-N · COVER | F1 score: 0.312 -> 0.437 | Helped |
| add image flipping augmentation (ablation) | RGB-N · NIST16 | F1 score: 0.716 -> 0.722 | No clear effect |
| add JPEG compression or Gaussian-noise augmentation beyond flipping (ablation) | RGB-N · NIST16, COVER, CASIA | F1 score: Flipping 0.722/0.437/0.408; JPEG 0.712/0.425/0.413; noise 0.717/0.412/0.396 | Hurt |
| use only the RGB stream instead of RGB and noise streams (ablation) | Faster R-CNN manipulation detector · Synthetic COCO test | AP: 0.445 -> 0.627 | Hurt |
| use only the noise stream instead of RGB and noise streams (ablation) | Faster R-CNN manipulation detector · Synthetic COCO test | AP: 0.461 -> 0.627 | Hurt |
Closest papers by meaning
- UnionFormer: Unified-Learning Transformer with Multi-View Representation for Image Manipulation Detection and Localization (CVPR 2024 · similarity 0.88)
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.86)
- IDseq: Decoupled and Sequentially Detecting and Grounding Multi-Modal Media Manipulation (AAAI 2025 · similarity 0.85)
- Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective (CVPR 2024 · similarity 0.85)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.84)
- A New Benchmark and Model for Challenging Image Manipulation Detection (AAAI 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.