← all datasets

CSTBIR

Emerging
1papers using it
2025first seen

CSTBIR (Composite Sketch+Text Based Image Retrieval) is a dataset containing approximately 2 million queries and 108,000 natural scene images, used to evaluate the retrieval of objects based on composite multimodal queries that combine hand-drawn sketches and descriptive text.

🔎 Find this dataset

Papers using CSTBIR (1)

CSTBIR — datasets — learning-to-hash