<![CDATA[Quality classification of local fibrous coconuts is still largely performed manually, a process that requires precision and significant time, while also potentially leading to inconsistencies in quality assessment. This study aims to develop an Android-based quality classification application for local fibrous coconuts using the Convolutional Neural Network (CNN) algorithm with the MobileNetV2 architecture. The research involved several stages: dataset collection (coconut images), preprocessing, data augmentation, model training, model testing, and implementation of the model into an Android application using TensorFlow Lite. The model was developed to classify coconut quality into three categories: immature, mature, and reject. The results demonstrate that the MobileNetV2 model effectively learned the visual characteristics of local fibrous coconuts, yielding optimal classification performance based on evaluations using accuracy, loss, precision, recall, F1-score, and a confusion matrix. The developed model was successfully implemented in an Android application, enabling automated classification using images captured via the device's camera or selected from its gallery. The findings indicate that the developed application offers a practical and efficient solution for the rapid and consistent quality identification of local fibrous coconuts. According to the research provided, the dataset consisted of 3,000 images of local fibrous coconuts, comprising 1,000 images each for the immature (semi-ripe), mature (ripe), and reject (damaged) classes. The dataset was split into 70% training data (2,100 images), 15% validation data (450 images), and 15% testing data (450 images). Following the training process using the CNN algorithm with the MobileNetV2 architecture, the model achieved a test accuracy of 96% and a test loss of 0.1098, demonstrating excellent classification performance in distinguishing between the three coconut quality classes.]]>
