Increasing traffic management complexity in the urban space presents opportunity for valid and timely vehicle detection systems. Many traditional surveillance methods do not perform well in low visibility or high throughput situations. This paper presents a pragmatic solution using the YOLOv8 object detection algorithm which is designed specifically for urban traffic environments. It is based upon a custom video dataset (ind1.mp4) derived from a real world traffic situation that replaces traditional datasets (i.e. COCO dataset). YOLOv8 is applied to each video frame to detect traffic movement. Its output detection, as cropped detections, is then further processed with the Gemini API for attribute characteristics, such as vehicle class, colour, manufacturer and license plate. The solution uses Python and OpenCV to provide a practical system which performs well in demonstrated formats, based upon accurate and reliable object detection and tracking. This also confirms the model's functional application for real time traffic applications such as traffic density, and control, as well as lane based tasks.