<![CDATA[Manual determination of papaya fruit ripeness is still widely practiced by farmers and traders, relying solely on visual observation. This approach is prone to errors caused by subjectivity and the limitations of human perception. This study aims to implement the Transfer Learning method using the DenseNet121 architecture to automatically classify papaya fruit ripeness levels based on digital images. The dataset used is divided into three ripeness classes: unripe, semi-ripe, and ripe. The research stages include dataset collection, image preprocessing using ImageDataGenerator, data division into training, validation, and testing sets, and model training using the Adam optimizer with a learning rate of 0.001, batch size of 32, and 50 epochs. The model was built using the Python programming language with the TensorFlow and Keras libraries. Furthermore, model performance was evaluated using a confusion matrix, classification report, and ROC curve. The results show that the DenseNet121 model was able to learn the visual characteristics of papaya ripeness levels very well. Based on the evaluation, the model achieved an accuracy of 97.96%, with high precision, recall, and F1-score values for each ripeness category. In addition, the ROC curve yielded an Area Under Curve (AUC) value close to 1.00, indicating excellent classification performance. These findings demonstrate that the application of Transfer Learning with the DenseNet121 architecture is effective in classifying papaya fruit ripeness levels. This method has the potential to be developed into a decision support system for farmers and agricultural businesses to improve the efficiency and accuracy of the fruit sorting process.]]>
