Waseda_Meisei_SoftBank at Video Browser Showdown 2024
摘要
This paper presents our first interactive video browser system “System 4 Vision”. Because our system is based on the system that achieved the highest video retrieval accuracy in the AVS task of the TRECVID benchmark 2022, high retrieval accuracy can be expected in the Video Browser Showdown competition. Our system is characterized by the availability of rich text input, including complicated multiple conditions as queries, because our system uses the visual-semantic embedding method represented by Contrastive Language-Image Pre-training (CLIP).