Research map / Image generation / Text-to-image diffusion
ImageRAG: Dynamic Image Retrieval for Reference-Guided Image Generation
Rotem Shalev Arkushin, Rinon Gal, Amit H. Bermano, Ohad Fried
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 ImageRAG retrieval | OmniGen · iNaturalist | GPTScore: 0.07 -> 0.56 | Improved |
| add ImageRAG retrieval | FLUX · iNaturalist | GPTScore: 0.07 -> 0.31 | Improved |
| add ImageRAG retrieval | SDXL · ImageNet | GPTScore: 0.86 -> 0.92 | Improved |
| add ImageRAG retrieval | SDXL · iNaturalist | CLIP similarity: 0.259 -> 0.243 | Hurt |
| add ImageRAG retrieval | SDXL · CUB | CLIP similarity: 0.315 -> 0.314 | No clear effect |
| add ImageRAG retrieval | OmniGen · iNaturalist | CLIP similarity: 0.155 -> 0.197 | Improved |
| add prompt rephrasing only (no references) (ablation) | OmniGen · ImageNet | CLIP similarity: 0.247 -> 0.248 | No clear effect |
| add detailed image captions for retrieval instead of missing concepts (ablation) | ImageRAG-O · ImageNet | CLIP similarity: 0.258 -> 0.264 | Improved |
| add detailed image captions for retrieval instead of the raw prompt (ablation) | ImageRAG-O · CUB | DINO similarity: 0.736 -> 0.760 | Improved |
| add detailed image captions for retrieval instead of missing concepts (ablation) | ImageRAG-O · CUB | DINO similarity: 0.719 -> 0.760 | Improved |
| use GPT re-ranking instead of CLIP retrieval (ablation) | ImageRAG-O · ImageNet | CLIP similarity: 0.264 -> 0.265 | No clear effect |
| use SigLIP instead of CLIP for retrieval (ablation) | ImageRAG-O · ImageNet | CLIP similarity: 0.264 -> 0.259 | 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.
- OmniGen — OmniGen: Unified Image Generation (CVPR 2025) · better on 5 of 5 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
- RealRAG: Retrieval-augmented Realistic Image Generation via Self-reflective Contrastive Learning (ICML 2025 · similarity 0.90)
- GenIR: Generative Visual Feedback for Mental Image Retrieval (NeurIPS 2025 · similarity 0.89)
- TIGeR: Unifying Text-to-Image Generation and Retrieval with Large Multimodal Models (ICLR 2025 · similarity 0.89)
- VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents (ICLR 2025 · similarity 0.88)
- Retrieval-Augmented Perception: High-resolution Image Perception Meets Visual RAG (ICML 2025 · similarity 0.88)
- AR-RAG: Autoregressive Retrieval Augmentation for Image Generation (NeurIPS 2025 · 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.
