Scholarlink

This study aims to develop a real-time classification system for two types of fish, Anabas testudineus (puyu) and Oreochromis niloticus (tilapia), in a controlled environment. Accurate and efficient classification of fish species is crucial for monitoring and managing aquatic ecosystems. Despite advances in object detection algorithms, there remains a need for comparative studies to identify the most effective models for specific applications. Three models of (You Only Look Once) YOLO object detection algorithms, YOLOv3, YOLOv4 Tiny, and YOLO (Neural Architecture Search) NAS, were utilized to achieve the most accurate classification results. The models were trained and tested on a new dataset of 1,643 images. Key metrics such as mean Average Precision (mAP), F1 score, recall, and precision were calculated to compare the performance of the models. The system, employing machine learning techniques, achieved y values of 93.4% with YOLOv3, 98.6% with YOLOv4 Tiny, and 68.2% with YOLO NAS at a 50% confidence threshold. These results highlight the varying effectiveness of each model in accurately classifying the fish species. The study concludes that YOLOv4 Tiny outperforms the other models in terms of accuracy and is the most suitable for real-time implementation in controlled environments. The findings provide valuable insights for selecting appropriate object detection algorithms in aquatic studies.

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

Classification of Anabas testudineus and Oreochromis niloticus Using Deep Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amir FakarulIsroq Abdul Razak,
  • Muhammad Nur Aiman Shapiee,
  • Mohd Izzat Mohd Rahman,
  • Nur Aliya Syahirah Badrol Hisam,
  • Muhammad Amirul Abdullah,
  • Mohd Azraai Mohd Razman

摘要

This study aims to develop a real-time classification system for two types of fish, Anabas testudineus (puyu) and Oreochromis niloticus (tilapia), in a controlled environment. Accurate and efficient classification of fish species is crucial for monitoring and managing aquatic ecosystems. Despite advances in object detection algorithms, there remains a need for comparative studies to identify the most effective models for specific applications. Three models of (You Only Look Once) YOLO object detection algorithms, YOLOv3, YOLOv4 Tiny, and YOLO (Neural Architecture Search) NAS, were utilized to achieve the most accurate classification results. The models were trained and tested on a new dataset of 1,643 images. Key metrics such as mean Average Precision (mAP), F1 score, recall, and precision were calculated to compare the performance of the models. The system, employing machine learning techniques, achieved y values of 93.4% with YOLOv3, 98.6% with YOLOv4 Tiny, and 68.2% with YOLO NAS at a 50% confidence threshold. These results highlight the varying effectiveness of each model in accurately classifying the fish species. The study concludes that YOLOv4 Tiny outperforms the other models in terms of accuracy and is the most suitable for real-time implementation in controlled environments. The findings provide valuable insights for selecting appropriate object detection algorithms in aquatic studies.