VIVA: Visual Exploration and Analysis of Videos with Interactive Annotation

This paper presents VIVA, a novel interactive tool for visually exploring long videos and searching for specific moments. Previous work on video data exploration and analytics often assumes that manually-created, rich annotations are available. However, such metadata may not be easily obtained. We design an interactively machine learning workflow for users to rapidly create annotations along a timeline. Combined with VIVA’s focus+context visualization that effectively displays frame snapshots in the context of a video stream, VIVA enables users to explore and analyze long video clips by incrementally make sense of them. We present usage scenarios that demonstrate how users would use VIVA for video-related tasks.