Research map / Vision-language and multimodal / Human and embodied modeling
PhotoFramer: Multi-modal Image Composition Instruction
Zhiyuan You, Ke Wang, He Zhang, Xin Cai, Jinjin Gu, Tianfan Xue, Chao Dong, Zhoutong Zhang
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 |
|---|---|---|---|
| finetune Bagel on the composition triplet dataset | Bagel · Shift task benchmark, GPT-5 win rate vs original | win rate (%): 27.61 -> 80.37 | Improved |
| finetune Bagel on the composition dataset (text-image consistency) | Bagel · Text-image guidance consistency, average | consistency score: 83.10 -> 92.02 | Improved |
| use PhotoFramer instead of open-source editing models (Qwen-Image-Edit) | image editing · Zoom-in task benchmark, GPT-5 win rate vs ground truth | win rate (%): 39.65 -> 67.24 | Improved |
| use PhotoFramer vs gpt-image-1 on view-change | image generation · View-change task, GPT-5 win rate vs original | win rate (%): 84.61 vs 82.07 | No clear effect |
| allow free viewpoint changes (view-change task) for open-source editors | Qwen-Image-Edit · View-change vs shift task, GPT-5 win rate vs original | win rate (%): 46.01 -> 70.65 | Improved |
| feed PhotoFramer text guidance into Qwen-Image-Edit (ablation) | Qwen-Image-Edit · text-guided editing | fidelity / composition | Hurt |
| feed text guidance into gpt-image-1 (ablation) | gpt-image-1 · text-guided editing | composition | Improved |
| train with text guidance instead of image pairs only (ablation) | Bagel · qualitative (Fig 11) | composition | Improved |
| train Bagel on text + image instead of finetuning Kontext on images only (ablation) | finetuned editor · qualitative (Fig 12) | composition | Improved |
| use auto prompt (shift and/or zoom-in) instead of single task prompt (ablation) | PhotoFramer · qualitative (Fig 9) | composition | Improved |
| train composition assessment 7B model with GRPO vs Qwen2.5-VL-32B | assessment model · GAIC composition assessment | SRCC: 0.195 -> 0.795 | Improved |
Later papers that compared with this method
1 later accepted paper reports PhotoFramer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- COMPASS: Grounding Composition-Intent Guidance in Unified Multimodal Models (ECCV 2026 · better on 1 of 1 shared result)
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.
- Q-Align — Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels (ICML 2024) · better on 3 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
- Can Machines Understand Composition? Dataset and Benchmark for Photographic Image Composition Embedding and Understanding (CVPR 2025 · similarity 0.88)
- Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping (CVPR 2026 · similarity 0.87)
- Enhancing Vision-Language Compositional Understanding with Multimodal Synthetic Data (CVPR 2025 · similarity 0.87)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.87)
- Localizing and Editing Knowledge In Text-to-Image Generative Models (ICLR 2024 · similarity 0.86)
- Learning Visual Composition through Improved Semantic Guidance (CVPR 2025 · 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.
