Advancing Video-Text Pretraining with Multi-View Captions

Benchmark Model Rank Results
zero-shot-video-retrieval-on-activitynetMVC-B (10M)–text-to-video R@1: 61.0
zero-shot-video-retrieval-on-activitynetMVC-B (5M)–text-to-video R@1: 58.6
zero-shot-video-retrieval-on-activitynetMVC-L (10M)–text-to-video R@1: 66.9
zero-shot-video-retrieval-on-activitynetMVC-L (5M)–text-to-video R@1: 63.9
zero-shot-video-retrieval-on-didemoMVC-B (10M)–text-to-video R@1: 61.0
zero-shot-video-retrieval-on-didemoMVC-B (5M)–text-to-video R@1: 58.9
zero-shot-video-retrieval-on-didemoMVC-L (10M)–text-to-video R@1: 62.0
zero-shot-video-retrieval-on-didemoMVC-L (5M)–text-to-video R@1: 61.2
zero-shot-video-retrieval-on-msr-vttMVC-B (10M)–text-to-video R@1: 39.0
zero-shot-video-retrieval-on-msr-vttMVC-B (5M)–text-to-video R@1: 38.2
zero-shot-video-retrieval-on-msr-vttMVC-L (10M)–text-to-video R@1: 41.1
zero-shot-video-retrieval-on-msr-vttMVC-L (5M)–text-to-video R@1: 40.6