MViewCon: A Multi-View Video Summarization Dataset

MViewCon is a custom-built multi-view surveillance video dataset developed to support academic research in query-based and multi-view video summarization. The dataset consists of synchronized video recordings captured from multiple CCTV cameras observing the same environment from different viewpoints under varying lighting and environmental conditions.

The dataset includes real-world indoor and outdoor scenes such as car parking areas, medical stores, jewellery shops, and electronic stores, enabling research on view relevance estimation, feature minimization, multi-camera synchronization, and scene understanding. Each video is accompanied by structured metadata (e.g., camera ID, location, recording date and time, duration, resolution, frame rate, bit rate, view angle, and lighting conditions).

The dataset contains video recordings of human subjects captured in semi-public and business-owned environments where video surveillance is routinely in operation. Informed consent was obtained from property and business owners prior to data collection, and informed consent was obtained from individuals appearing in the recordings, with the option to withdraw participation. Footage containing individuals who did not provide consent has been excluded.

The study received approval from the Institutional Ethics Committee , under Registration Number: EC/NEW/IND/2024/MH/0444.

This dataset is available under controlled access and may only be used after the approval and acceptance of the Dataset Usage Agreement (DUA). The dataset is intended solely for academic research purposes, and commercial use of the video footage is not permitted. Researchers seeking access must request the Data Use Agreement (DUA) by contacting [e-mail]. Upon receiving the DUA, applicants must complete and sign the agreement and return it via email. Access will be granted to qualified researchers upon review and approval of the signed agreement.