This study presents a lightweight Convolutional Neural Network (CNN) model for the binary classification of apple leaf images into healthy and Alternaria diseased categories, aimed at enhancing orchard management practices through automated disease detection. A self-collected dataset of 600 high-resolution apple leaf images, balanced between healthy and diseased samples, was acquired under real-world conditions and annotated by domain experts. The dataset underwent preprocessing, including color-based enhancement to highlight diseased regions, and was augmented with transformations to improve model generalization. The proposed CNN architecture, comprising three convolutional blocks and a total of 110,785 parameters, was optimized for computational efficiency and realtime applications. The model employed the Adam optimizer during training, with a specified learning rate of 0.0001 and regularization techniques to prevent overfitting. The Evaluation metrics, accuracy, recall, precision, and F1-score, demonstrated robust performance, achieving an overall accuracy of 89.17%, a Recall of 89.17%, a Precision of 84.20%, and an F1-score of 86.54%. The training and validation curves signify stable learning dynamics and minimal overfitting, while the confusion matrix further confirms the model’s strong classification capabilities. This study underscores the potential of the proposed CNN model for scalable and reliable apple leaf disease detection, contributing to sustainable agricultural practices.