← all datasets

FSIR-BD

Emerging
1papers using it
2026first seen

The FSIR-BD dataset is a benchmark containing 38,353 images and 303 queries designed to evaluate few-shot text-to-image retrieval, specifically targeting compositional and out-of-distribution queries with reference examples.

🔎 Find this dataset

Papers using FSIR-BD (1)

FSIR-BD — datasets — learning-to-hash