Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/168782
Title: Peningkatan Efisiensi Penghitungan Benih Ikan Hias Melalui Otomatisasi Berbasis Kecerdasan Buatan
Other Titles: 
Authors: Syafutra, Heriyanto
Alauddin, Muhammad Fakhri
Issue Date: 2025
Publisher: IPB University
Abstract: Penghitungan benih ikan hias di Balai Benih Ikan Kota Sukabumi hingga kini masih banyak dilakukan secara manual dan kurang efisien karena memakan waktu lama, rentan kesalahan, serta bergantung pada ketelitian petugas. Untuk mengatasi hal ini, penelitian ini mengembangkan sistem otomatis berbasis kecerdasan buatan menggunakan algoritma deteksi objek YOLOv5, dilatih dengan dataset gambar benih ikan melalui platform Roboflow. Sistem ini dijalankan pada mini PC Orange Pi yang terhubung dengan kamera serta antarmuka GUI untuk menampilkan hasil penghitungan secara real-time. Hasil pengujian menunjukkan performa baik dengan mAP@0.5 sebesar 96,9%, precision 96,9%, dan recall 92,5%. Sistem ini juga mampu meningkatkan efisiensi waktu penghitungan lebih dari 90% dibanding metode manual, dengan tingkat kesalahan deteksi yang rendah meskipun jumlah benih banyak. Dengan kinerja tersebut, sistem dinilai efektif sebagai alternatif metode manual, mendukung efisiensi proses pembenihan, dan telah mengadopsi pendekatan Internet of Things (IoT) melalui integrasi perangkat keras dan perangkat lunak, selaras dengan transformasi digital di sektor perikanan.
The counting of ornamental fish seeds at the Sukabumi City Fish Seed Center is still largely done manually and is inefficient because it is time-consuming, prone to errors, and dependent on the accuracy of the staff. To address this issue, this study developed an automated system based on artificial intelligence using the YOLOv5 object detection algorithm, trained with a dataset of fish seed images through the Roboflow platform. The system runs on an Orange Pi mini PC connected to a camera and a GUI interface to display real-time counting results. Testing results show good performance with a mAP@0.5 of 96.9%, precision of 96.9%, and recall of 92.5%. The system also improves counting efficiency by over 90% compared to manual methods, with low detection error rates even when dealing with large numbers of fish larvae. With such performance, the system is deemed effective as an alternative to manual methods, supporting the efficiency of the fish breeding process, and has adopted an Internet of Things (IoT) approach through the integration of hardware and software, aligning with digital transformation in the fisheries sector.
URI: http://repository.ipb.ac.id/handle/123456789/168782
Appears in Collections:UT - Computer Engineering Tehcnology

Files in This Item:
File Description SizeFormat 
cover_J0304211121_068e14c601244cdab6b113e3a12e7162.pdfCover2.2 MBAdobe PDFView/Open
fulltext_J0304211121_7ae21d07f111447aa112be81c9555925.pdf
  Restricted Access
Fulltext10 MBAdobe PDFView/Open
lampiran_J0304211121_eae41bca5d814ac5b3fea6687ae8843a.pdf
  Restricted Access
Lampiran2.3 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.