Comparison of the Latest Version of Deep Learning YOLO Model in Automatically Detecting the Ripeness Level of Oil Palm Fruit in PTPN IV, North Sumatra, Indonesia

This research explores the capability of the Deep Learning models YOLOv8 and YOLOv9 in detecting the ripeness level of oil palm fruit. The purpose of this research is to compare the accuracy results between YOLOv8 and YOLOv9 in detecting the ripeness level of oil palm fruit and to find out whether there is an improvement in the accuracy of YOLOv9 compared to the previous version, YOLOv8. This research uses datasets sourced from the oil palm plantation repository of PTPN IV Medan, North Sumatra, Indonesia. The dataset collected for this research consists of 180 images with a size of 640x640 and a camera resolution of 12 Mega Pixel. The dataset was uploaded into Roboflow and underwent pre-processing and augmentation. After obtaining the API from Roboflow and inputting it into Google Colab for YOLOv8 and YOLOv9 training, testing was carried out on the provided images. The results show that both YOLOv8 and YOLOv9 successfully detected ripe, rotten, and unripe oil palm fruit with an accuracy of 0.984 (98.4^%) and 0.99 (99^%). The results show that YOLOv9 has higher accuracy than YOLOv8 in detecting the ripeness level of oil palm fruit. The potential use of the latest version of the YOLO model in detecting the ripeness level of oil palm fruit in plantations can serve as a basis for developing more effective and efficient automatic detection systems in the future.