Skripsi
IMPLEMENTASI CONVOLUTIONAL NEURAL NETWORK DENGAN ARSITEKTUR MOBILENETV3 PADA APLIKASI ANDROID UNTUK KLASIFIKASI TINGKAT KEMATANGAN TOMAT
Tomatoes are a horticultural commodity whose quality and nutritional content are strongly influenced by their ripeness level. In practice, tomatoes are often sorted manually based on ripeness, which leads to inconsistent results due to the limitations of human visual perception. This problem highlights the need for an objective and fast tomato ripeness classification system. This study aims to develop an Android application that is able to classify the ripeness level of tomatoes by utilizing a Convolutional Neural Network (CNN) based on the MobileNetV3 architecture implemented in an Android application. The software development method used in this research is Extreme Programming (XP). The ripeness levels used as categories in this study are ripe, half-ripe, and unripe. The dataset consists of 1,350 tomato images, with a data split of 80% for training, 10% for validation, and 10% for testing. The training results show an average accuracy of 84%, precision of 86%, recall of 84%, and F1-score of 84%. Furthermore, based on the black-box testing performed, all test scenarios succeeded as expected.