Research map / Image generation / Text-to-image diffusion
Watch Your Steps: Local Image and Scene Editing by Text Instructions
Ashkan Mirzaei, Tristan T Aumentado-Armstrong, Marcus A Brubaker, Jonathan Kelly, Alex Levinshtein, Konstantinos G Derpanis, Igor Gilitschenski
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 relevance-map mask guidance to IP2P image editing | InstructPix2Pix · IP2P test set (5,000 images) | CLIP image similarity vs text-image direction similarity | Improved |
| raise mask threshold tau (ablation) | relevance-guided image editing · IP2P test set | image similarity and edit success | Mixed |
| raise image guidance scale in IP2P instead of using relevance masks (ablation) | InstructPix2Pix · IP2P test set | text-image direction similarity | Hurt |
| add relevance field and relevance-guided edits to iterative NeRF dataset updates | IN2N · NeRF editing, 14 tasks | Edit PSNR: 28.70 -> 31.01 | Improved |
| add relevance field and relevance-guided edits to iterative NeRF dataset updates | IN2N · NeRF editing, 14 tasks | NIQE: 6.43 -> 5.53 | Improved |
| add relevance-guided localization | IN2N · NeRF editing, 14 tasks | CLIP text-image similarity: 0.2683 -> 0.2673 | No clear effect |
| edit NeRF views with IP2P independently per frame (ablation) | IP2P · NeRF editing, 14 tasks | CLIP frame similarity: 0.9876 -> 0.9669 (ours vs per-frame IP2P) | Hurt |
| add relevance field versus IN2N on CLIP edit similarity | IN2N · NeRF editing, 14 tasks | CLIP edit similarity: 0.8822 -> 0.8754 | No clear effect |
| compute relevance at high noise level t_rel = 0.8 (ablation) | relevance map · relevance maps | mask quality (qualitative) | Improved |
| supervise a relevance field from multiple views instead of single-view maps (ablation) | relevance map · relevance maps | mask accuracy (qualitative) | Improved |
Closest papers by meaning
- Editable Noise Map Inversion: Encoding Target-image into Noise For High-Fidelity Image Manipulation (ICML 2025 · similarity 0.86)
- Noise Map Guidance: Inversion with Spatial Context for Real Image Editing (ICLR 2024 · similarity 0.86)
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.86)
- Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference (AAAI 2024 · similarity 0.86)
- ZONE: Zero-Shot Instruction-Guided Local Editing (CVPR 2024 · similarity 0.86)
- Localizing and Editing Knowledge In Text-to-Image Generative Models (ICLR 2024 · similarity 0.85)
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.
