A Computer Vision System to Classify the Quality of Central Sulawesi Cocoa Beans Using Convolutional Neural Networks (CNN)

The cocoa industry serves as a vital economic pillar, holding a pivotal role in the daily lives of farmers in Central Sulawesi. However, a prevailing challenge for cocoa cultivators in this region resides in the inherently subjective nature of cocoa bean quality assessment, often necessitating the involvement of a substantial human workforce. Addressing this challenge requires the implementation of computer vision technology capable of autonomously grading cocoa bean quality. Such a system not only expedites the evaluation process but also enhances the precision and consistency in determining the quality of the beans. The study employed Convolutional Neural Networks (CNN) methodology to analyze images of cocoa beans sourced from Central Sulawesi farmers. The research demonstrated a remarkable accuracy of 94.66%, accompanied by a minimal loss of 0.1195. The model demonstrates strong performance on both training and validation sets, indicating effective learning of relevant patterns. Consistent accuracy and loss trends across epochs signify stable training without notable overfitting or underfitting. The minimal gap between training and validation metrics reflects the model's effective generalization to unseen data. This technological integration promises to revolutionize cacao quality assessment, offering a sophisticated and efficient solution for the benefit of Central Sulawesi's cacao farming community.