Ulla Delfana Rosiani, Nurlaily Asrobika, Daffa Setya Nugraha, Mungki Astiningrum, Mamluatul Hani'ah, Gunawan Budiprasetyo
Accurate and efficient enumeration of catfish seed is essential for the aquaculture industry to meet increasing market demand. Conventional manual counting methodologies can induce stress, injuries, and illnesses in fish, potentially compromising their growth and wellbeing. This study aimed to develop an automated tool for quantifying and categorizing catfish seed according to their size by utilizing YOLOv8 and MobileNetV2 to enhance precision and efficiency. The system captures fry movements using OpenMV Cam H7 Plus, followed by detection, tracking, size classification, and enumeration. According to the Indonesian National Standard (SNI: 01-6484.2-2000), catfish seed fall into three size categories: Grade A (1-3 cm), Grade B (3-5 cm), and Grade C (5-8 cm). To optimize the accuracy of this process, two object-detection techniques were evaluated: the Blob method and YOLOv8. These methods, combined with the deep learning capabilities of YOLOv8 and MobileNetV2, ensure that the fry are accurately counted and classified. The experimental results demonstrated that the integration of YOLOv8 with MobileNetV2 yielded an average accuracy of 90.32%. The precision of this system renders it an ideal solution for real-time catfish seed detection and size categorization, significantly improving the operational efficiency and reducing manual errors for aqua culturists. © 2024 IEEE.
Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia