Multi-camera Large-Scale Intelligent Video Analytics

The Multi-camera Large-Scale Intelligent Video Analytics project aims to develop a comprehensive system for analyzing videos captured by multiple cameras in an environment. This field involves a wide range of tasks, including real-time processing, edge computing, object detection and tracking, identification and recognition, analytics and machine learning, and scalability. The challenge is to solve these tasks in this project while maintaining scalability. Therefore, this project utilizes the Omniverse to capture desired videos from various perspectives and generate synthetic data. The synthetic data are trained on a YOLOv8 model using the Ultralytics framework. The model is deployed using DeepStream, a powerful video analytics framework, which incorporates techniques such as ROI filtering, object detection, tracking, and direction detection. The system allows for the identifying and tracking of specific objects of interest within defined regions, providing valuable insights into their presence, location, and movement. Furthermore, the project explores the application of these techniques for inventory management and safety monitoring, demonstrating their effectiveness in enhancing warehouse operations and ensuring a secure environment. The results of this project contribute to the development of intelligent video analytics systems that can be utilized in various real-world scenarios, improving efficiency, and enabling informed decision-making.