MViewCon: A Multi-View Video Summarization Dataset

MViewCon is a custom-built multi-view surveillance video dataset developed for research in query-based video summarization. The dataset includes video recordings from multiple cameras capturing the same environment from different angles, under varied lighting and view conditions. It supports tasks involving query-based summarization, feature selection, view relevance estimation, and multi-view understanding. The accompanying metadata file (metadata.xlsx) provides structured information about each video, such as camera ID, location, date/time, duration, resolution, frame rate, bit rate, view angle, and lighting conditions.

The dataset also includes annotations (annotations.rar) for specific video events, a videos.rar archive containing the raw multi-view footage, and a README.txt file with usage instructions, licensing information, and citation guidelines. MViewCon was used in a PhD project focused on feature minimization using symmetric uncertainty, and camera view selection using Prim’s algorithm. It is particularly useful for researchers and developers working on video summarization, surveillance analytics, multi-camera systems, and computer vision tasks requiring real-time or query-driven summarization capabilities.