<![CDATA[Abstrak - Penyakit daun anggur menjadi salah satu faktor utama yang menurunkan produktivitas dan kualitas tanaman anggur. Identifikasi manual yang dilakukan petani sering kali tidak akurat karena kemiripan gejala antar penyakit, sehingga diperlukan sistem otomatis yang cepat dan andal. Penelitian ini bertujuan mengembangkan sistem identifikasi penyakit daun anggur berbasis Convolutional Neural Network (CNN) dengan pendekatan pengolahan citra digital. Tiga arsitektur CNN dirancang dan dievaluasi menggunakan dataset PlantVillage yang dibagi menjadi 80% data latih dan 20% data validasi. Hasil pengujian menunjukkan Arsitektur 3 mencapai akurasi tertinggi sebesar 98%, diikuti Arsitektur 1 sebesar 97%, sementara Arsitektur 2 hanya 86%. Modifikasi dilakukan pada Arsitektur 2 dengan menambah jumlah filter konvolusi serta menghapus dropout, sehingga akurasinya meningkat signifikan menjadi 97%. Peningkatan terbesar terjadi pada kelas Healthy dari 62% menjadi 98%. Temuan ini membuktikan bahwa modifikasi arsitektur CNN mampu meningkatkan kinerja identifikasi penyakit daun anggur secara efektif.Kata kunci: CNN; daun anggur; identifikasi; pengolahan citra; penyakit tanaman; Abstract - Grape leaf diseases are one of the main factors that reduce the productivity and quality of grape plants. Manual identification by farmers is often inaccurate due to the visual similarity between symptoms, thus requiring a fast and reliable automated system. This study aims to develop a grape leaf disease identification system based on Convolutional Neural Networks (CNN) using a digital image processing approach. Three CNN architectures were designed and evaluated using the PlantVillage dataset, which was divided into 80% training data and 20% validation data. The experimental results show that Architecture 3 achieved the highest accuracy of 98%, followed by Architecture 1 with 97%, while Architecture 2 reached only 86%. Modifications were then applied to Architecture 2 by increasing the number of convolutional filters and removing the dropout layer, which significantly improved its accuracy to 97%. The most notable improvement occurred in the Healthy class, where accuracy increased from 62% to 98%. These findings demonstrate that modifying CNN architectures can effectively enhance the performance of grape leaf disease identification.Keywords: CNN; grape leaf; identification; image processing; plant disease;]]>
