Performance Analysis of Various Algorithms In 2D Dynamic Object Detection

Technology advancement produced the uncrewed autonomous ground vehicle. The multi-view real-time environmental state provision for autonomous uncrewed ground vehicles has recently gained popularity. In autonomous driving, object detection techniques will be used more and more. The most prominent 2D dynamic object detection algorithms are YOLO v5, YOLO v4, and Mask R-CNN. This work analyses the performance of these three algorithms on two datasets: COCO Dataset and NuScenes Dataset. Finally, the performance obtained using these models on these two different datasets is compared using the FPS, inference time, and mean average precision as the performance evaluation parameters. On analysing the performance, it can be seen that the YOLO v5 model performs better than the other two models in real-time scenarios. Therefore, in Intelligent Transport Systems, the YOLO v5 model is suitable for 2D dynamic object detection.