| NIM | L1C022078 |
|---|
| Namamhs | MARIO NABHAN MATNI AL-JUHDI |
|---|
| Judul Artikel | RANCANG BANGUN AUTOMATIC FISH COUNTER BERBASIS COMPUTER VISION |
|---|
| Abstrak (Bhs. Indonesia) | Penggunaan metode konvensional berbasis observasi visual dan pencacahan manual untuk estimasi jumlah ikan oleh tenaga manusia menunjukkan beberapa keterbatasan signifikan. Penelitian ini bertujuan untuk mengembangkan prototipe automatic fish counter berbasis computer vision untuk mendeteksi dan menghitung benih ikan lele secara otomatis dan real time. Prototipe dikembangkan menggunakan algoritma YOLOv8 untuk deteksi objek, metode vertical line cross detection untuk penghitungan objek, dan algoritma Simple Online and Realtime Tracking (SORT) untuk pelacakan objek. Model dilatih menggunakan 5055 citra, diuji menggunakan 241 citra, dan divalidasi menggunakan 481 citra. Metrik evaluasi menunjukkan nilai accuracy sebesar 87,70%, precision 94,52%, recall 92,40%, F1-score 93,45%, mAP@0.5 91,90%, dan mAP@0.5:0.95 48,78% sehingga menunjukkan kemampuan model dalam mendeteksi benih ikan lele dengan baik. Prototipe mampu mencapai nilai overall accuracy sebesar 92.17%, relative bias sebesar 7,83%, dan RMSE sebesar 5,25 pada proses penghitungan benih ikan lele dalam sejumlah rangkaian pengujian yang telah dilakukan. Hasil penelitian menunjukkan bahwa prototipe yang dikembangkan memiliki performa yang baik dan layak digunakan untuk melakukan penghitungan benih ikan lele secara otomatis dan real time. |
|---|
| Abtrak (Bhs. Inggris) | The use of conventional methods based on visual observation and manual counting to estimate fish numbers by human operators has several significant limitations. This study aims to develop a prototype automatic fish counter based on computer vision to detect and count catfish fry automatically and in real time. The prototype was developed using the YOLOv8 algorithm for object detection, the vertical line cross detection method for object counting, and the Simple Online and Real-time Tracking (SORT) algorithm for object tracking. The model was trained using 5055 images, tested using 241 images, and validated using 481 images. The evaluation metrics show an accuracy of 87.70%, precision of 94.52%, recall of 92.40%, an F1-score of 93.45%, a mAP@0.5 of 91.90%, and a mAP@0.5:0.95 of 48.78%, demonstrating the model’s ability to detect catfish fry effectively. The prototype achieved an overall accuracy of 92.17%, a relative bias of 7.83%, and an RMSE of 5.25 when counting catfish fry in a series of tests that were carried out. The research results indicate that the prototype developed performs well and is suitable for automatically counting catfish fry in real time. |
|---|
| Kata kunci | benih ikan lele, computer vision, fish counter, object detection, YOLOv8 |
|---|
| Pembimbing 1 | Rizqi Rizaldi Hidayat, S.I.K., M.Si. |
|---|
| Pembimbing 2 | Agung Cahyo Setyawan, S.Pi., M.Si. |
|---|
| Pembimbing 3 | |
|---|
| Tahun | 2026 |
|---|
| Jumlah Halaman | 103 |
|---|
| Tgl. Entri | 2026-07-31 10:53:37.417015 |
|---|