Fine-tuning BLIP-2 with a Q-Former raises FashionIQ validation Recall@10 to roughly 45%, but the zero-shot framing is inaccurate and the DPO variant omits the reference image.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Zero Shot Composed Image Retrieval
Fine-tuning BLIP-2 with a Q-Former raises FashionIQ validation Recall@10 to roughly 45%, but the zero-shot framing is inaccurate and the DPO variant omits the reference image.