Ripeness Classification of Fresh Oil Palm Using Deep Learning

The challenge of accurately ripeness level classifying from fresh oil palm fruit, this research investigated the problems and methods that were used in ripeness level classification. The goal of this research is to present a method for classifying the ripeness levels of oil palm fruits using deep learning techniques applied to image data. This research applies a deep learning model for classifying the ripeness level of fresh oil palm fruit and aiming for real-time application or mobile usage. The proposed model, Yolov9n (nano), adapts principles from convolutional neural networks that was tuned the learning parameters consist of the Adam optimizer, a learning rate of 0.001, and a batch size of 32, result in the most accurate ripeness level classification model. The research demonstrates the Yolov9n (nano) improved model with optimized parameters to achieve the highest accuracy. Training and validation datasets represent precision of 99.4%, recall of 99.6%, mean average precision at 50% (mAP50) of 99.5%, and mAP50-95 of 92.6%. These findings confirm the model’s effectiveness in accurately classifying the ripeness level of oil palm fruit from images. The study provides guidance for the practical deep learning implementation of ripeness level classification on fresh oil palm imaging for efficient harvesting by farmers cultivating oil palm.