Towards The Future of Crab Farming: The Application Of AI with Yolox And Yolov9 To Detect Crab Larvae

Crabs are a highly valued food source and a key export commodity for Indonesia, but farming them remains a challenge, particularly during the larval stage, where survival rates are critically low. This issue has contributed to the declining population of wild crabs. A crucial factor in improving survival rates is the accurate detection and counting of crab larvae. By determining the precise number of larvae, farmers can optimize feeding ratios, manage stocking densities to reduce cannibalism, maintain water quality, and improve cost efficiency through better resource management. Despite its importance, no affordable and precise tools currently exist for this purpose. This study aims to develop a cost-effective and accurate crab larvae detection and counting application using image processing powered by deep learning artificial intelligence (AI). Two models, You Only Look Once eXtreme (YOLOX) and YOLOv9, were evaluated for their performance. The YOLOX-S model struggled with accuracy in detecting larvae, whereas the YOLOv9 model demonstrated superior performance, achieving a mean Average Precision (mAP) of 0.85 at IoU=0.5 and successfully detecting 93% of crab larvae objects accurately. The findings of this research have significant implications for supporting Indonesia's blue economy and aligning with Sustainable Development Goals (SDGs), particularly in sustainable fisheries and aquaculture. By enabling a tech-driven approach to crab farming, this solution addresses the challenges of declining wild crab populations, improves food security, and promotes economic growth for fishers and farmers. These advancements contribute to the development of a sustainable crab farming ecosystem, ensuring long-term ecological and economic benefits.