Frozen in Time: A Joint Video and Image Encoder for End-to-End Retrieval

Benchmark Model Rank Results
video-retrieval-on-didemoFROZEN#30text-to-video R@1: 31.0text-to-video R@5: 59.8
video-retrieval-on-lsmdcFROZEN#23text-to-video R@1: 15.0text-to-video R@5: 30.8
video-retrieval-on-msr-vttFROZEN#16text-to-video R@1: 32.5text-to-video R@5: 61.5
video-retrieval-on-msr-vtt-1kaFROZEN#41text-to-video R@1: 31.0text-to-video R@5: 59.5
video-retrieval-on-msvdFROZEN#18text-to-video R@1: 33.7text-to-video R@5: 64.7
video-retrieval-on-querydFrozen#4text-to-video R@1: 53.8text-to-video R@5: 75.7
zero-shot-video-retrieval-on-didemoFROZEN#20text-to-video R@1: 21.1text-to-video R@5: 46.0
zero-shot-video-retrieval-on-didemoM. Bain et. al.#21text-to-video R@1: 20.2text-to-video R@5: 46.4
zero-shot-video-retrieval-on-msr-vttFROZEN#26text-to-video R@1: 24.7text-to-video R@5: 46.9