SPRIGHT

홈페이지 · 논문 1편

SPRIGHT is the first, large-scale vision-language dataset that focuses on spatial relationships. It contains ~6M images that have been re-captioned with a synthetic focus.