Electronic Theses and Dissertation

ABSTRAK
Aceh Tengah dikenal sebagai salah satu daerah utama penghasil kopi di Indonesia, yang terkenal dengan kopi Arabika Gayo. Meski demikian, daerah ini menghadapi tantangan besar dari serangan penyakit daun kopi seperti leaf rust, sooty mold, brown eye spot, dan black rot, yang dapat menurunkan kualitas dan produksi kopi. Selama ini, identifikasi penyakit daun kopi masih dilakukan secara konvensional yang bergantung pada pengetahuan dan pengalaman petani, serta rentan terhadap kesalahan dan membutuhkan waktu yang lama. Penelitian ini bertujuan untuk membangun model deep learning untuk klasifikasi penyakit daun kopi Arabika Gayo berdasarkan citra daun, sehingga dapat membantu petani dalam mendeteksi penyakit secara cepat dan akurat. Model dikembangkan menggunakan arsitektur EfficientNet-B0, ResNet50, dan VGG16, dan dilatih pada dataset primer yang terdiri dari 1.200 citra daun kopi yang terbagi secara merata ke dalam empat kelas penyakit. Citra dikumpulkan dari tiga lokasi perkebunan di Kecamatan Jagong Jeget, Aceh Tengah, dan data dibagi ke dalam set pelatihan, validasi, dan pengujian dengan rasio 80:10:10. Model terbaik diperoleh dari arsitektur EfficientNet-B0 pada konfigurasi learning rate 0,01 dan 60 epoch, dengan accuracy 97,50%, precision 97,58%, recall 97,50%, dan F1-score 97,50%. Hasil ini menunjukkan bahwa pendekatan deep learning, khususnya menggunakan arsitektur EfficientNet-B0, efektif dalam mengklasifikasikan penyakit daun kopi dan dapat menjadi solusi pendeteksian penyakit yang lebih efisien dibandingkan metode konvensional.

Kata Kunci: Penyakit daun kopi, convolutional neural network (CNN), klasifikasi citra, EfficientNet-B0, ResNet50, VGG16

ABSTRACT Aceh Tengah is known as one of the main coffee-producing regions in Indonesia, particularly famous for its Gayo Arabica coffee. However, this region faces major challenges from coffee leaf diseases such as leaf rust, sooty mold, brown eye spot, and black rot, which can significantly reduce coffee quality and yield. Traditionally, disease identification has relied on farmers’ knowledge and experience, which is prone to error and time-consuming due to the need for manual inspection of leaf symptoms. This study aims to develop a deep learning model to classify diseases on Gayo Arabica coffee leaves based on leaf images, thereby assisting farmers in detecting diseases quickly and accurately. The models were built using three convolutional neural network architectures: EfficientNet-B0, ResNet50, and VGG16. The dataset consists of 1,200 primary images, evenly distributed into four disease classes. The leaf images were collected from three coffee plantations in Jagong Jeget subdistrict, Central Aceh, and the data were split into training, validation, and testing sets in an 80:10:10 ratio. The best performance was achieved using the EfficientNet-B0 architecture with a learning rate of 0.01 and 60 epochs, resulting in an accuracy of 97.50%, precision of 97.58%, recall of 97.50%, and F1-score of 97.50%. These results indicate that deep learning approaches, particularly EfficientNet-B0, are effective for classifying coffee leaf diseases and offer a more efficient alternative to conventional identification methods. Keywords: Coffee leaf disease, convolutional neural network (CNN), image classification, EfficientNet-B0, ResNet50, VGG16