Research map / Image generation / Editing and style transfer
RAA: Achieving Interactive Remove/Add Anything via Fully Synthetic Data
Delong Liu, Haotian Hou, Zhaohui Hou, Shihao Han, Zhiyuan Huang, Mingjie Zhan, Fei Su, Zhicheng Zhao
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 ROI bounding box input via mask-input highlight (ablation) | RAA · MagicBrush (add/remove subset) | GPT pass rate: 0.701 -> 0.794 | Improved |
| condition on ROI via mask-input rather than cross-attention embedding (ablation) | RAA · MagicBrush (add/remove subset) | GPT pass rate: 0.734 -> 0.794 | Improved |
| raise LoRA rank from 32 to 64 (ablation) | RAA · MagicBrush (add/remove subset) | GPT pass rate: 0.794 -> 0.775 | No clear effect |
| multi-round self-improving fine-tuning of FLUX.1-dev for diptychs (ablation) | RAD pipeline · RAD diptych synthesis | pass rate: 0.056 -> 0.773 | Improved |
| additional self-improvement iterations beyond the third (ablation) | RAD pipeline · RAD diptych synthesis | pass rate: 0.699 -> 0.773 | No clear effect |
| RAA versus FLUX.1-Kontext with instruction only | RAA · MagicBrush (add/remove subset) | GPT pass rate: 0.682 -> 0.701 | Improved |
| RAA versus FLUX.1-Kontext with instruction only | RAA · Emu Edit (add/remove subset) | GPT pass rate: 0.621 -> 0.665 | Improved |
| mask-input methods versus RAA on preservation metrics | Flux-Fill · MagicBrush (add/remove subset) | L1: 0.042 vs 0.049 (RAA) | Mixed |
Closest papers by meaning
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · similarity 0.90)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.88)
- Object-Aware Inversion and Reassembly for Image Editing (ICLR 2024 · similarity 0.87)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.86)
- FunEditor: Achieving Complex Image Edits via Function Aggregation with Diffusion Models (AAAI 2025 · similarity 0.86)
- UltraEdit: Instruction-based Fine-Grained Image Editing at Scale (NeurIPS 2024 · similarity 0.86)
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.
