Comparison of YOLO Algorithm Performance in Classifying Harvest Periods for Pleurotus Ostreatus

Pleurotus ostreatus is an economically important crop with high market demand. The challenge in harvesting Pleurotus ostreatus lies in ensuring the produce is harvested at the right time to maintain high quality. However, farmers often lack the knowledge and equipment to accurately determine the optimal harvest timing, relying heavily on personal experience. This study aims to address this issue by employing the You Only Look Once (YOLO) algorithm, known for its speed and accuracy, to create a model capable of classifying suitable harvest periods. The research conducted performance evaluations and accuracy comparisons among various YOLO model versions: YOLOv5n, YOLOv5s, YOLOv7-tiny, YOLOv7, YOLOv8n, and YOLOv8s. A custom dataset specific to Pleurotus ostreatus was utilized, segmented into two harvest periods: the preparation-to-harvest period and the ready-to-harvest period. Experimental results demonstrate that YOLOv8s achieved the highest accuracy at 95.40%, whereas YOLOv5n, the smallest model, achieved the lowest accuracy at 84.48%. This research is significant as it contributes to the development of technology aimed at enhancing the efficiency of Pleurotus ostreatus harvesting practices. It provides a promising avenue for agricultural practitioners to achieve sustainable production.