Artificial Intelligence-Driven Video Indexing for Rapid Surveillance Footage Summarization and Review

Artificial Intelligence-Driven Video Indexing for Rapid Surveillance Footage Summarization and Review

Jaemin Jung, Soonyong Park, Harim Kim, Changha Lee, Charmgil Hong

Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence
Demo Track. Pages 8687-8690. https://doi.org/10.24963/ijcai.2024/1009

This paper introduces VIDEX, an advanced tool designed to streamline the analysis of surveillance video through a user-friendly interface. VIDEX achieves high development efficiency and maintainability utilizing the Model-View-ViewModel (MVVM) design pattern. The core feature of VIDEX is a footage summary using object detection and anomaly detection. Its architecture ensures efficient data management by organizing detected objects and anomalies within an indexed database, thus facilitating a more rapid review process. Additionally, multi-threading was used to shorten the processing time. VIDEX provides a video summarization that can be used primarily in the criminal investigation stage using the information stored in a database. Discover more about VIDEX and access its resources at https://github.com/nth221/videx.
Keywords:
Machine Learning: ML: Applications
Computer Vision: CV: Recognition (object detection, categorization)
Data Mining: DM: Anomaly/outlier detection