Video Summarization

Abstract

In today's society, a lot of information is released every day in various formats, including text, video, and images. Day to day YouTube is generating large amount of data. In July 2015, YouTube revealed that it receives over 400 h of video content every single minute, which translates to 65.7 years. worth of content uploaded every day. Since then, we are experiencing an even stronger engagement of consumers with both online video platforms and devices (e.g., smartphones and wearables) that carry powerful video recording sensors and allow instant uploading of the captured video on the Web Video summarization technologies aim to create a concise and complete synopsis by selecting the most informative parts of the video content. Generally, LSTM is used for this video summarization. In this project, we present an automated system for extracting specific person's footage from large surveillance videos. The system employs computer vision techniques, including person detection and recognition algorithms, to analyse each frame of the video and identify the desired persons of interest. The person detection algorithm utilizes the YOLOv3 object detection model to locate persons in the frames, while the person recognition algorithm utilizes a pre-trained face recognition model to verify the identity of the detected persons. By combining these algorithms, the system identifies frames that contain the target persons and extracts them for further analysis. The extracted frames are then used to create a short video comprising the selected footage. The system offers a convenient and efficient solution for surveillance video analysis, allowing for the isolation of relevant footage and reducing the need for manual inspection.