New framework PeFuse enables training-free Composed Image Retrieval

Researchers have developed PeFuse, a novel training-free framework for Composed Image Retrieval (CIR). This method utilizes pre-trained Diffusion Models and Multimodal Large Language Models to bridge different modalities through generative conversion. PeFuse reformulates CIR into single-modality retrieval tasks, bypassing the need for specialized training and achieving competitive performance on standard benchmarks. AI

IMPACT This research could improve the efficiency and flexibility of image search systems by leveraging existing large models without additional training.

RANK_REASON The cluster describes a research paper detailing a new framework for image retrieval. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX
• Composed Image Retrieval
• CORE Recommender
• DagsHub
• Diffusion Models
• Gotit.pub
• Hugging Face
• Multimodal Large Language Models
• PeFuse
• ScienceCast

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →