Image-based classification of herbal plant leaves plays a vital role in supporting agricultural digitization and rapidmedicinal plant identification. This study aims to develop and analyze a classification system for papaya (Carica papaya)and betel (Piper betle) leaves using a Convolutional Neural Network (CNN) and a hybrid CNN–Support Vector Machine(SVM) approach. The dataset, obtained from a public Kaggle repository, consisted of 1,000 images equally dividedbetween the two classes and was split into 90% for training and 10% for validation. Each image was resized to 150×150pixels and augmented to enrich data variability. The CNN model, trained for up to 30 epochs using the Adam optimizer,achieved a training accuracy of 99.82% and a validation accuracy of 47%, indicating overfitting caused by its 3,824,934parameters. In contrast, the CNN–SVM hybrid model—using CNN as a feature extractor and SVM as a classifier—achieved a validation accuracy of 89% with balanced precision and recall across both classes. These findings demonstratethat while CNN effectively captures visual features, integrating SVM enhances generalization on small datasets. Thisresearch contributes to the development of efficient, stable, and interpretable herbal plant classification systems based onneural networks