Efficient Real-Time Vehicle Tracking Using YOLOv8 with Frame Skipping

Tracking vehicles in real-time traffic videos is crucial for traffic analysis and violation detection. However, processing every frame in a 30 FPS video using YOLOv8 for vehicle detection significantly increases computational costs. To address this, we propose a fixed frame skipping approach, where detection is performed on 10 frames per second instead of all 30 frames, which are then fed into the DeepSeek tracking algorithm to ensure continuous tracking. To prevent data loss from skipped frames, our method approximates bounding boxes for those frames. This approach reduces computational overhead while maintaining tracking accuracy, making it efficient for real-time traffic surveillance