<![CDATA[Cherry tomatoes, small in size but sweet and nutritious, are increasingly popular and cultivated in Indonesia. The main challenge is determining optimal ripeness for harvesting. Technologies such as YOLOv8 can improve the accuracy of fruit ripeness detection. This study at BBPP Lembang applied YOLOv8 to detect cherry tomato ripeness and analyzed the model's performance and influencing factors. The research method for cherry tomato ripeness analysis using YOLOv8 involved image collection and annotation, data augmentation, and dataset sharing. The model was trained on the Google Colab platform with GPUs, evaluated using metrics such as accuracy and mAP, and then integrated into a website. The system was tested in real-world conditions and monitored regularly to ensure optimal performance. YOLOv8 successfully identified cherry tomato ripeness at BBPP Lembang with a mAP of 77.5%, a precision of 77.6%, a recall of 68.1%, and an F1-score of 72.5% after 100 epochs. The model demonstrated 85.71% accuracy in field trials, succeeding in 6 out of 7 tests. Model performance is influenced by dataset quality, lighting, and image resolution for optimal results. The model achieved 85.71% accuracy in real-time conditions on the website. Dataset quality and lighting influence performance, demonstrating YOLOv8's effectiveness in accurately monitoring fruit ripeness]]>
Copyrights © 2025
