Research map / Image generation / Text-to-image diffusion
Edit2Perceive: Image Editing Diffusion Models Are Strong Dense Perceivers
Yiqing Shi, Yiren Song, Mike Zheng Shou
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 image-editing base (FLUX.1 Kontext) instead of text-to-image FLUX.1 (ablation) | Edit2Perceive depth, basic config · NYUv2 | AbsRel: 6.8 -> 5.1 | Improved |
| use image-editing base instead of T2I base with full recipe (ablation) | Edit2Perceive depth · KITTI | AbsRel: 8.4 -> 7.9 | Improved |
| use image-editing base instead of T2I for normals (ablation) | Edit2Perceive normals · NYUv2 normals | mean angular error: 16.4 -> 15.7 | Improved |
| add pixel-space consistency loss (ablation) | T2I FLUX.1 depth · NYUv2 | AbsRel: 6.8 -> 5.4 | Improved |
| add pixel-space consistency loss (ablation) | FLUX.1 Kontext depth · NYUv2 | AbsRel: 4.7 -> 4.4 | Improved |
| add pixel-space consistency loss (ablation) | FLUX.1 Kontext depth, uniform mapping · KITTI | AbsRel: 9.6 -> 9.6 | No clear effect |
| add pixel-space consistency loss (ablation) | FLUX.1 Kontext normals · NYUv2 normals | 11.25 deg accuracy: 60.0 -> 61.6 | Improved |
| use sqrt depth mapping instead of uniform (ablation) | FLUX.1 Kontext depth, no LCons · KITTI | AbsRel: 9.6 -> 8.2 | Improved |
| use sqrt depth mapping instead of uniform (ablation) | T2I FLUX.1 depth, no LCons · KITTI | AbsRel: 13.2 -> 10.2 | Improved |
| increase inference steps beyond 4 (ablation) | Edit2Perceive · depth and normal benchmarks | delta1 / 11.25 deg accuracy | Hurt |
| single-step inference instead of multi-step (ablation) | Edit2Perceive · depth and normal benchmarks | accuracy | No clear effect |
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.
- SDMatte — SDMatte: Grafting Diffusion Models for Interactive Matting (ICCV 2025) · better on 2 of 3 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
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.90)
- PAIR Diffusion: A Comprehensive Multimodal Object-Level Image Editor (CVPR 2024 · similarity 0.90)
- High-Fidelity Diffusion-Based Image Editing (AAAI 2024 · similarity 0.90)
- PixPerfect: Seamless Latent Diffusion Local Editing with Discriminative Pixel-Space Refinement (NeurIPS 2025 · similarity 0.89)
- PIXELS: Progressive Image Xemplar-based Editing with Latent Surgery (AAAI 2025 · similarity 0.89)
- ConsistDreamer: 3D-Consistent 2D Diffusion for High-Fidelity Scene Editing (CVPR 2024 · similarity 0.89)
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.
