Research map / Detection and segmentation / Open-set and few-shot detection
Generate, Segment, and Refine: Towards Generic Manipulation Segmentation
Peng Zhou, Bor-Chun Chen, Xintong Han, Mahyar Najibi, Abhinav Shrivastava, Ser-Nam Lim, Larry Davis
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 GSR-Net rather than the strongest reported baseline | Manipulation segmentation · Carvalho, In-The-Wild, COVER, CASIA | F1: Carvalho: 0.520 -> 0.525; In-The-Wild: 0.424 -> 0.555; COVER: 0.276 -> 0.489; CASIA: 0.541 -> 0.574 | Improved |
| add simple copy-pasted training images (ablation) | DeepLab VGG-16 · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.420 -> 0.446; 0.472 -> 0.504; 0.376 -> 0.410; 0.474 -> 0.503 | Improved |
| use generated training images instead of simple copy-pasted images (ablation) | DeepLab VGG-16 · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.446 -> 0.460; 0.504 -> 0.524; 0.410 -> 0.434; 0.503 -> 0.506 | Improved |
| add both copy-pasted and generated training images (ablation) | DeepLab VGG-16 · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.460 -> 0.472; 0.524 -> 0.528; 0.434 -> 0.444; 0.506 -> 0.507 | Improved |
| use proposed generated images instead of deep harmonization-generated images (ablation) | DeepLab VGG-16 · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.384 -> 0.460; 0.421 -> 0.524; 0.342 -> 0.434; 0.420 -> 0.506 | Improved |
| add boundary-guided segmentation (ablation) | DeepLab + copy-pasted and generated training images · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.472 -> 0.515; 0.528 -> 0.540; 0.444 -> 0.455; 0.507 -> 0.545 | Improved |
| add the refinement stage that creates new boundary artifacts (ablation) | GS-Net · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.515 -> 0.525; 0.540 -> 0.555; 0.455 -> 0.489; 0.545 -> 0.574 | Improved |
| add generated images to copy-pasted training images (ablation) | Segmentation trained using COCO annotations · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.343 -> 0.354; 0.430 -> 0.441; 0.351 -> 0.355; 0.242 -> 0.270 | Improved |
| add the refinement stage (ablation) | Segmentation trained using COCO annotations and copy-pasted plus generated images · Carvalho, In-The-Wild, COVER, CASIA | F1: 0.354 -> 0.418; 0.441 -> 0.479; 0.355 -> 0.381; 0.270 -> 0.331 | Improved |
Closest papers by meaning
- Proactive Image Manipulation Detection (CVPR 2022 · similarity 0.88)
- Image Manipulation Detection by Multi-View Multi-Scale Supervision (ICCV 2021 · similarity 0.87)
- IDseq: Decoupled and Sequentially Detecting and Grounding Multi-Modal Media Manipulation (AAAI 2025 · similarity 0.85)
- GIM: A Million-scale Benchmark for Generative Image Manipulation Detection and Localization (AAAI 2025 · similarity 0.84)
- IMDPrompter: Adapting SAM to Image Manipulation Detection by Cross-View Automated Prompt Learning (ICLR 2025 · similarity 0.84)
- Generative Photomontage (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.
