Bird.s Eye: Analysis of Video Surveillance Data using Deep Learning

Video surveillance data is widely used as a measure for security in various places like residential areas, shopping malls and so on. This video data is recorded 24x7 and therefore the task to find a particular object or event becomes tedious. Precious manual hours are wasted by security officials trying to skim through the video data trying to find the object they are searching for. This paper aims to reduce the manual effort required in this process by automating and focusing on the object the user is searching for. We have implemented multi-class object detection using YOLO algorithm and deep neural networks. The application will detect commonplace objects, classify them accordingly and extract the features of these objects. The user will be able to query the video data using the name of the object and the features. The paper is expected to detect objects of interest and extract basic features with reasonable accuracy and processing times through which it can be deployed in real-world surveillance applications.