Kepiting bakau merupakan komoditas perikanan yang memiliki nilai ekonomi tinggi, namun identifikasi gender kepiting ini masih dilakukan secara manual, yang memakan waktu dan berpotensi menyebabkan stres pada kepiting. Penelitian ini bertujuan untuk membangun model deteksi gender kepiting bakau secara otomatis menggunakan algoritma YOLO. Metode penelitian yang digunakan mengikuti kerangka kerja CRISP DM. Model YOLO dilatih untuk mendeteksi gender berdasarkan karakteristik fisik abdomen kepiting. Hasil penelitian menunjukkan bahwa YOLO11n memberikan performa terbaik pada split data 80:10:10 dengan precision 99,8%, recall 100%, F1-Score 100%, accuracy 100%, mAP@50 sebesar 99,5%, dan mAP@50-95 sebesar 84,3%, sedangkan YOLOv8n unggul pada split 70:20:10 dengan precision 99,8%, recall 100%, F1-Score 100%, accuracy 100%, mAP@50 sebesar 99,5%, dan mAP@50-95 sebesar 82,8%. Temuan ini menunjukkan bahwa pemilihan model terbaik bergantung pada proporsi data, sehingga implementasi model YOLO11n dan YOLOv8n dapat menjadi solusi efektif dalam mendeteksi gender kepiting bakau secara otomatis berbasis citra, serta mendukung efisiensi operasional dalam industri perikanan.

Mangrove crabs are a fishery commodity with high economic value; however, gender identification is still performed manually, which is time-consuming and may cause stress to the crabs. This study aims to develop an automated gender detection model for mangrove crabs using the YOLO algorithm. The research method followed the CRISP DM framework. The YOLO model was trained to detect gender based on the physical characteristics of the crab's abdomen. The results show that YOLO11n achieved the best performance on the 80:10:10 data split with a precision of 99.8%, recall of 100%, F1 Score of 100%, accuracy of 100%, mAP@50 of 99.5%, and mAP@50-95 of 84.3%, while YOLOv8n performed best on the 70:20:10 split with a precision of 99.8%, recall of 100%, F1-Score of 100%, accuracy of 100%, mAP@50 of 99.5%, and mAP@50-95 of 82.8%. These findings indicate that the selection of the best model depends on the data split proportion; therefore, the implementation of YOLO11n and YOLOv8n models can serve as an effective solution for automatic mangrove crab gender detection based on images, thereby supporting operational efficiency in the fisheries industry.