by Jingzhou LiuResearch Only
by Jingzhou LiuLicense : Research Only
A new large-scale dataset, named VIOLIN (VIdeO-and-Language INference), is introduced for this task, which consists of 95,322 videohypothesis pairs from 15,887 video clips, spanning over 582 hours of video. These video clips contain rich content with diverse temporal dynamics, event shifts, and people interactions, collected from two sources: (i) popular TV shows, and (ii) movie clips from YouTube channels. Video-and-Language Inference, for joint multimodal understanding of video and text. Given a video clip with aligned subtitles as premise, paired with a natural language hypothesis based on the video content, a model needs to infer whether the hypothesis is entailed or contradicted by the given video clip.
TasksVideo-and-Language Inference, Video Understanding, Text Understanding