Research map / Image generation / Editing and style transfer
FreeEdit: Mask-Free Reference-Based Image Editing With Multi-Modal Instruction
Runze He, Kai Ma, Linjiang Huang, Shaofei Huang, Jialin Gao, Xiaoming Wei, Jiao Dai, Jizhong Han, Si Liu
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 Replace-type data to Add-only training (ablation) | FreeEdit · Reference object addition | CLIP-R: 75.97 -> 77.02 | Improved |
| add Remove-type data (no reference image) (ablation) | FreeEdit · Reference object addition | CLIP-IQA: 0.655 -> 0.705 | Improved |
| add instruction recaptioning (ablation) | FreeEdit · Reference-based editing ablation | DINO-I: 51.54 -> 53.38 | Improved |
| add instruction recaptioning (ablation) | FreeEdit · Reference-based editing ablation | DINO-R: 49.44 -> 48.74 | Hurt |
| unfreeze and train the Q-Former (ablation) | FreeEdit · Reference-based editing ablation | DINO-R: 48.74 -> 51.26 | Improved |
| unfreeze and train the Q-Former (ablation) | FreeEdit · Reference-based editing ablation | CLIP-I: 77.95 -> 75.17 | Hurt |
| add decoupled residual refer-attention (DRRA) over no refer-attention (ablation) | FreeEdit · Reference-based editing ablation | DINO-R: 51.26 -> 56.71 | Improved |
| use DRRA instead of RASA (ablation) | FreeEdit · Refer-attention ablation | DINO-I: 51.77 -> 52.29 | Improved |
| add quality tuning stage (ablation) | FreeEdit · Reference-based editing ablation | CLIP-I: 76.95 -> 81.59 | Improved |
| add quality tuning stage (ablation) | FreeEdit · Reference-based editing ablation | DINO-R: 56.71 -> 54.37 | Hurt |
| increase reference scale lambda (ablation) | FreeEdit · Reference-based editing | similarity to reference | Improved |
| remove mask requirement (FreeEdit vs mask-based MimicBrush) | FreeEdit · Reference-based editing benchmark | CLIP-R: 74.99 -> 78.12 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- InstructPix2Pix — InstructPix2Pix: Learning To Follow Image Editing Instructions (CVPR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- InsightEdit: Towards Better Instruction Following for Image Editing (CVPR 2025 · similarity 0.90)
- FireEdit: Fine-grained Instruction-based Image Editing via Region-aware Vision Language Model (CVPR 2025 · similarity 0.89)
- AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea (CVPR 2025 · similarity 0.89)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.88)
- UltraEdit: Instruction-based Fine-Grained Image Editing at Scale (NeurIPS 2024 · similarity 0.88)
- Referring Image Editing: Object-level Image Editing via Referring Expressions (CVPR 2024 · similarity 0.88)
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.
