Klasifikasi Citra Biji Kopi Sangrai Arabika dan Robusta Menggunakan Convolutional Neural Network

Al Firdaus, Muhammad Rafi
Mardhiyyah, Rodhiyah
Sanjaya, Fadil Indra

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 7 (2025): December 2025

Coffee is one of Indonesia's leading commodities, with two main varieties: Arabica and Robusta. The differences in characteristics between these two types of coffee, such as bean shape, color, and texture, are often difficult to distinguish visually, especially for the general public. This study aims to develop an automatic classification system capable of distinguishing Arabica and Robusta coffee beans using the Convolutional Neural Network (CNN) method with the application of transfer learning based on the MobileNetV2 architecture. The dataset used consists of 210 images of coffee beans taken using a smartphone camera with various positions and lighting, which were then divided into training data (60%), validation data (20%) and test data (20%). Before the training process, data augmentation such as rotation, zoom, flip, and brightness adjustment was performed to enrich image variation and reduce the risk of overfitting. Training was conducted with a learning rate of 0.0001, a batch size of 32, and an Adam optimizer. The results showed that the CNN model with MobileNetV2 transfer learning was able to achieve a training accuracy of 99.21% and a testing accuracy of 97.62%, with relatively low loss values of 0.0682 for training data and 0.1333 for validation data. The application of transfer learning contributes to improving the stability of the training process by utilizing the pre-trained weights from the ImageNet model. Based on these results, it can be concluded that the MobileN-based CNN method.