CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retrieval

Benchmark Model Rank Results
video-retrieval-on-activitynetCLIP4Clip#23text-to-video R@1: 40.5text-to-video R@5: 73.4
video-retrieval-on-didemoCLIP4Clip#28text-to-video R@1: 43.4text-to-video R@5: 70.2
video-retrieval-on-lsmdcCLIP4Clip#20text-to-video R@1: 21.6text-to-video R@5: 41.8
video-retrieval-on-msr-vttCLIP4Clip-seqTransf#10text-to-video R@1: 44.5text-to-video R@5: 71.4
video-retrieval-on-msr-vtt-1kaCLIP4Clip#51text-to-video R@10: 81.6text-to-video Median Rank: 2
video-retrieval-on-msvdCLIP4Clip#15text-to-video R@1: 46.2text-to-video R@5: 76.1
zero-shot-video-retrieval-on-lsmdcCLIP4Clip#10text-to-video R@1: 15.1text-to-video R@5: 28.5
zero-shot-video-retrieval-on-msr-vttCLIP4Clip#19text-to-video R@1: 32.0text-to-video R@5: 57.0
zero-shot-video-retrieval-on-msvdCLIP4Clip#12text-to-video R@1: 38.5text-to-video R@5: 66.9