Advanced Classification of Oil Palm Fruit Ripeness Using ResNet50 and Real-Time Image Analysis for Enhanced Agricultural Practices

The classification of oil palm fruit ripeness is a critical factor in optimizing palm oil production. Traditional methods of ripeness assessment, based on the percentage of detached fruits and changes in skin color, are prone to human error due to subjective judgment. This study proposes an advanced approach utilizing deep learning with the ResNet50 model to classify oil palm fruit ripeness into four levels: unripe, under-ripe, ripe, and overripe. The research evaluates the model's performance under various data allocations, optimizers, and learning rates while incorporating data augmentation techniques to enhance accuracy. Experimental results indicate the optimal configuration includes a 90/10 data split, Adam optimizer, and a learning rate of 0.0001, achieving precision of 96%, recall of 98%, F1 score of 97%, and accuracy of 97%. These findings highlight the potential of ResNet50 in delivering reliable, real-time classification for agricultural applications, providing a practical solution for farmers and industries. The study concludes that large and diverse training datasets are essential for achieving robust classification results