Kualitas dan tingkat kematangan buah pisang sangat memengaruhi harga jual dan kepuasan konsumen, namun proses penentuan secara manual memerlukan waktu, keahlian khusus, dan rawan kesalahan, sehingga dibutuhkan sistem otomatis yang akurat dan efisien. penelitian ini bertujuan merancang dan membangun sistem klasifikasi otomatis tingkat kematangan dan jenis buah pisang berbasis citra dengan menggunakan metode convolutional neural network (cnn), yang diharapkan dapat membantu petani, pedagang, dan konsumen dalam menilai kualitas buah secara cepat dan akurat. dataset yang digunakan mencakup 900 citra awal dari pisang jenis susu dan muli dengan tiga kategori kematangan yaitu mentah, matang, dan busuk. setelah melalui tahap preprocessing jumlah dataset meningkat menjadi 2.700 citra. selanjutnya, data dibagi menjadi data latih 70%, data validasi 15%, dan data uji 15%. model cnn dilatih menggunakan arsitektur vgg16 dan dievaluasi dengan metrik akurasi, precision, recall, dan f1-score. proses analisis data meliputi pembagian dataset, pelatihan model, serta evaluasi performa dengan mengamati kestabilan nilai metrik pada data validasi dan pengujian. hasil pengujian dari data uji dalam dataset menunjukkan performa model yang tinggi dengan nilai rata-rata precision 85%, recall 88%, dan f1-score 86% serta rata – rata akurasi 87,90%. hasil pengujian sebanyak 405 data dari data uji diluar dataset menunjukkan performa model yang turun dengan nilai rata-rata precision 86%, recall 77% dan f1-score 79% serta rata – rata akurasi 76,79%. dengan demikian, sistem ini terbukti efektif sebagai sistem otomatis dalam penilaian kualitas buah pisang, dan dapat diimplementasikan dalam industri pertanian untuk meningkatkan efisiensi, keakuratan, serta mendukung pengambilan keputusan secara cepat dan otomatis. kata kunci: klasifikasi buah pisang, cnn, tingkat kematangan, citra digital.

The quality and ripeness of bananas have a significant impact on the selling price and customer satisfaction. however, the manual determination process is time consuming, specialized, and prone to errors. consequently, there is a need for an accurate and efficient automated system. the purpose of this research is to desig and build an automatic classification system for banana ripeness levels and types of bananas based on image by utilizing the convolutional neural network (cnn) method, which is expected to help farmers, traders, and consumers in rapidly and accurately assessing fruit quality. the dataset includes 900 initial images of milk and muli bananas, which are classified into three categories based on their ripeness unripe, ripe, and rotten. following the completion of the preprocessing stage, the number of datasets increased to 2,700 images. furthermore, the data was divided into 70% training data, 15% validation data, and 15% test data. the cnn model was trained using the vgg 16 architecture and evaluated with accuracy, precision, recall, and fi-score metrics. the data analysis process involves the sharing of datasets, the training of models, and the evaluation of performance through the observation of the stability of metric values in both validation and test data. test results from test data in the dataset show high model performance with an average value of precision 85%, recall 88%, and f1-score 86% and an average accuracy of 87,90%. the test results of 405 data from test data outside the dataset show that the model performance has decreased with an average value of precision 86%, recall 77% and fi-score 79% and an average accuracy of 76.79%. thus, this system is proven to be effective as an automated system in banana fruit quality assessment. it can be implemented in the agricultural industry to improve efficiency, accuracy, and support decision making quickly and automatically. keywords: banana classification, cnn, ripeness level, digital image