Agriculture is one of the most important, vital and commercial sectors for sustaining global food supply. However, potato diseases significantly threaten crops yield, quantity and quality, often resulting in a huge of farmers and food insecurity. Early and accurate disease detection is essential for timely precaution. In this study, we propose a hybrid and efficient deep learning model that integrates Convolutional Neural Networks (CNNs) for localized feature extraction with Transformer-based attention mechanisms for capturing long-range dependencies in leaf images of potato. The model is trained on a diverse and large dataset of potato leaves with multiple disease categories. Through rigorous evaluation, our approach demonstrates improved accuracy, robustness against variations in lighting and background, and better generalization to unseen samples compared to traditional CNN-only methods. Additionally, we provide an in-depth analysis of performance metrics, including confusion matrices, precision, recall, and F1-score, to validate the model’s reliability for real-world agricultural applications.