IMPLEMENTASI ALGORITMA YOLO11 DALAM MENDETEKSI SPESIES IKAN LAUT KOMERSIAL SECARA REAL TIME UNTUK SISTEM PENYORTIRAN IKAN

IMPLEMENTATION OF YOLO11 ALGORITHM TO DETECT COMMERCIAL MARINE FISH SPECIES IN REAL TIME FOR FISH SORTING SYSTEM

Laut Indonesia memiliki potensi komersial yang sangat besar, tak dapat dipungkiri bahwa lautan Indonesia memiliki sumber daya ikan yang cukup kaya. Proses penyortiran ikan berdasarkan spesiesnnya akan menjadi tantangan besar pada saat proses pemasaran atau distribusi, karena masih dilakukan secara manual. Penelitian ini mengimplementasikan algoritma model YOLO11 dalam mendeteksi spesies ikan laut untuk membantu pross penyortiran. Data yang digunakan terdiri dari 4 sumber dataset yang berbeda, yaitu Final Dataset, Detection Fish Computer Vision Project, Fish Tun Computer Vision Project, dan Fish-gress Dataset for Fish Species Classification. Dataset tersebut kemudian diolah menjadi dataset baru melalui proses data pre-processing dan data augmentation. Model YOLO11 yang digunakan pada saat proses pelatihan adalah YOLO11n dan YOLO11s, dengan 3 variasi pembagian data. Adapun  3 variasi pembagian data yang digunakan adalah 70% data training, 20% data validation, dan 10% data testing; 80% data training, 10% data validation, 10% data testing; dan 60% data training, 20% data validation, dan 20% data testing. Hasil pelatihan terbaik dihasilkan oleh model YOLO11n dengan pembagian data 70% data training, 20% data validation, dan 10% data testing. Model pelatihan dengan hasil terbaik menghasilkan nilai mAP 95.1%, precision 93.2%, dan recall 91.0%, memnunjukkan bahwa model menunjukkan performa yang tinggi. Penelitian ini membuka peluang untuk pengembangan sistem deteksi spesies ikan dalam membantu proses penyortiran.

   Kata Kunci – YOLO11, Deteksi Spesies Ikan Laut Komersial, Visi Komputer

The Indonesian sea has enormous commercial potential, it cannot be denied that the Indonesian sea has rich fish resources. The process of sorting fish based on their species will be a big challenge during the marketing or distribution process, because the sorting is handled manually. This research implements the YOLO11 model algorithm in detecting marine fish species to help the sorting process. The data used consists of 4 different dataset sources, namely Final Dataset, Detection Fish Computer Vision Project, Fish Tun Computer Vision Project, and Fish-gress Dataset for Fish Species Classification. These datasets are then processed into new datasets through data pre-processing and data augmentation. The YOLO11 models used during the training process are YOLO11n and YOLO11s, with 3 variations of data splitting. The 3 variations of data splits used are 70% training data, 20% validation data, and 10% testing data; 80% training data, 10% validation data, 10% testing data; and 60% training data, 20% validation data, and 20% testing data. The best training results are generated by the YOLO11n model with data split of 70% training data, 20% validation data, and 10% testing data. The training model with the best results achieved a mAP value of 95.1%, precision 93.2%, and recall 91.0%, indicating that the model showed high performance. This research opens up opportunities for the development of fish species detection systems to assist the sorting process.

               Keywords - YOLO11, Commercial Marine Fish Species Detection, Computer Vision