Scholarlink

Fish detection is crucial in aquaculture as it enables continuous monitoring of marine species. Traditional machine learning methods, although widely used, have limitations, particularly when dealing with the complexity of the underwater environment. Our approach combines the bounding boxes generated by YOLOV5 and Faster R-CNN to provide more accurate detection, with bounding boxes better adjusted to the objects. By weighting confidence scores, we optimize the confidence rate, thereby better reflecting the newly fused bounding boxes. The results demonstrated the effectiveness of our approach compared to simple detection using YOLOV5 or Faster R-CNN alone.

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

An Optimized Bounding Box Technique for Enhanced Underwater Fish Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Mahdi Hamzaoui,
  • Mohamed Ould-Elhassen Aoueileyine,
  • Seifeddine Bouallegue,
  • Ridha Bouallegue

摘要

Fish detection is crucial in aquaculture as it enables continuous monitoring of marine species. Traditional machine learning methods, although widely used, have limitations, particularly when dealing with the complexity of the underwater environment. Our approach combines the bounding boxes generated by YOLOV5 and Faster R-CNN to provide more accurate detection, with bounding boxes better adjusted to the objects. By weighting confidence scores, we optimize the confidence rate, thereby better reflecting the newly fused bounding boxes. The results demonstrated the effectiveness of our approach compared to simple detection using YOLOV5 or Faster R-CNN alone.