IP-CIR creates imagined proxy images from a query image and caption via LLM-based layout and conditional generation, then blends proxy, query, and text features to improve zero-shot composed image retrieval.
Conditioned and composed image retrieval combining and partially fine-tuning clip-based features
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Imagine and Seek: Improving Composed Image Retrieval with an Imagined Proxy
IP-CIR creates imagined proxy images from a query image and caption via LLM-based layout and conditional generation, then blends proxy, query, and text features to improve zero-shot composed image retrieval.