Real-Time Object Detection and Geolocation System Using Computer Vision

Abstract

Real-time, web-based object detection and tracking system that integrates state of the art in deep learning models with geolocation mapping. The system, which uses YOLOv8 and SSD for object detection, leverages the accuracy and speed of Convolutional Neural Networks (CNNs). The detected objects are tracked across video frames using the DeepSORT algorithm, which ensures consistent identity assignment by leveraging appearance descriptors and tracking motion. Built using TensorFlow and OpenCV, the system captures video from a standard webcam and processes frames in real time. Also, the Google Maps JavaScript API allows determining the location of an object and the camera's position from the map. The modular design separates detection and tracking processes, allowing for flexibility and scalability. This work demonstrates a practical and accessible approach to intelligent object monitoring, combining computer vision and geospatial technologies within a browser-based interface.