Scholarlink

Road maintenance and safety are important activities and potholes detection that require accurate and efficient detection methods. In this work, we propose a YOLOv5-based pothole detection model and compare its performance with yolov3 and yolov8. Our dataset consists of 665 scaled images of 640 × 640 pixels, and the models were trained for 100 epochs. YOLOv5m output mAP (mean average precision) 0.927 (50% IoU), 0.913 (IoU force) 50–95% understandably better. Its performance in accuracy and inference speed was better than YOLOv3 and YOLOv8, with overall precision of 0.792 and a recall of 0.736. Due to its efficient design and impressive accuracy, the YOLOv5m model serves as an excellent candidate for real-time pothole detection, leading to improved road safety and maintenance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative Analysis of Object Detection Algorithms for Pothole Detection Under Environmental Constraints

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sumaiya Mahdiya Mahia,
  • Faisal Imran,
  • Abdullah Al Yousuf Khan,
  • Md. Fazlah Karim Alvee,
  • Md. Shaifur Rahman,
  • Md. Sabbir Hossain

摘要

Road maintenance and safety are important activities and potholes detection that require accurate and efficient detection methods. In this work, we propose a YOLOv5-based pothole detection model and compare its performance with yolov3 and yolov8. Our dataset consists of 665 scaled images of 640 × 640 pixels, and the models were trained for 100 epochs. YOLOv5m output mAP (mean average precision) 0.927 (50% IoU), 0.913 (IoU force) 50–95% understandably better. Its performance in accuracy and inference speed was better than YOLOv3 and YOLOv8, with overall precision of 0.792 and a recall of 0.736. Due to its efficient design and impressive accuracy, the YOLOv5m model serves as an excellent candidate for real-time pothole detection, leading to improved road safety and maintenance.