Object Detection in Unstructured Driving Environments

This paper conducts a comprehensive error analysis of the inference process performed on the YOLOv8 and RTDETR model, utilizing two distinct datasets: MS COCO, on which YOLOv8 and RT-DETR is originally trained, and IDD, a separate dataset. The primary focus lies on evaluating model performance using mean Average Precision (mAP) and Intersection over Union (IoU) metrics. Through rigorous experimentation and analysis, we investigate the discrepancies in model performance when applied to these diverse datasets. The findings shed light on the strengths and weaknesses of the YOLOv8 and RT-DETR model across different data domains, offering valuable insights for improving object detection systems in real-world applications.