Get Started
Home
Topics
Search
Library
Research questionHow can composed image retrieval distinguish changed, preserved, and removed visual details?A whole-image similarity score can mix the edit with details that should remain or disappear. Retrieval therefore needs to represent different roles of the reference and text without losing fine-grained evidence.
AI
Computer Vision
Information Retrieval
Machine Learning
Multimodal Models
Latest papersRecent research connected to this question, newest first.MulVec: Fine-Grained Role-Aware Matching for Training-Free Zero-Shot Composed Image RetrievalThe source presents a training-free, zero-shot approach using frozen encoders, global and local visual representations, and role-specific query signals to rank gallery images. Evidence is reported on CIRCO, CIRR, and FashionIQ with three backbone scales, so applicability is limited to those retrieval settings.research paper · Sep 4, 2026
Related questions
How can reference-guided image generation control distinct attributes of multiple objects in complex scenes?How can generative image systems preserve subject identity under viewpoint changes, degradation, and iterative edits?How can image-text retrieval focus on caption-described attributes while ignoring unmentioned visual information?How can text-guided image editing explain edits while keeping provenance resistant to white-box removal?