Classification of Jengkol (Archidendron Pauciflorum) Varieties using Deep Learning

Archidendron pauciflorum, also known in Indonesia as jengkol, stands as a favored delicacy in the Southeast Asian region. The various varieties of jengkol exhibit remarkably high similarities, posing a challenge in their differentiation and potentially leading to economic losses. This research aims to develop a model capable of classifying various jengkol variants. The study employs state-of-the-art convolutional neural network architectures, VGG-16 and Inception V3, with and without pre-trained weights. A comprehensive analysis is conducted using a carefully curated dataset, and the performance of the models is evaluated based on accuracy, precision, recall, and F1 score metrics. The results indicate that Inception V3, particularly when using pre-trained weights, outperforms VGG-16, achieving an accuracy of 84.5%, whereas the accuracy of VGG-16 is only 77.5%. The research sheds light on the practical applications of accurate jengkol classification in culinary, agricultural, and ecological contexts. The findings highlight the potential of deep learning for enhancing the precision and reliability of jengkol variety identification in various real-world scenarios.