Color-Based Ripeness Classification of Oil Palm Fresh Fruit Bunches Using Convolution Neural Networks and Data Augmentation

Proper ripeness categorization of oil palm fresh fruit bunches is essential to ensure they are harvested at the optimal stage to maximize oil output. The current manual method of ripeness assessment, which relies on harvesters visually inspecting the fruit's color, is both labor-intensive and prone to human error. To address these limitations, there is a growing need for an automated classification system that leverages artificial intelligence. The manual classification process is cumbersome and susceptible to inaccuracies due to its reliance on subjective visual inspection. This labor-intensive approach increases the likelihood of human error, underscoring the necessity for a more reliable and efficient automated solution. Thus, this study aims to evaluate the feasibility of using Convolutional Neural Networks for automating the grading based on color. A dataset was compiled from various sources and augmented using a data generator. The model was constructed incorporating a transfer learning approach with an Xception model. The model achieved an overall accuracy of 80.4%. Visualization of the model using activation maps confirmed that the model can effectively classify the color, indicating its potential viability for the task.