SYNTH-PEDES

홈페이지 · 논문 3편

SYNTH-PEDES is a large-scale person dataset with image-text pairs by far, which contains 312,321 identities, 4,791,711 images, and 12,138,157 textual descriptions. Source: PLIP: Language-Image Pre-training for Person Representation Learning Source: PLIP | Github Repo