One of the most extensively grown crops in the world, potatoes are essential to food security. However, potato production is severely affected by several leaf diseases, including Late Blight, Early Blight, Mosaic Virus (PVY), and Black Leg. Conventional diagnostic techniques rely on manual inspection, which is labour-intensive, prone to error, and unsuitable for extensive farming regions. In this paper, a Convolution Neural Network (CNN)-based automated potato leaf disease detection system based on deep learning and image processing is presented. A labeled dataset of five classes (Late Blight, Early Blight, Mosaic Virus, Black Leg, and Healthy) was collected and pre-processed using advanced augmentation and contrast-enhancement techniques. The proposed hybrid CNN achieved an overall accuracy of 95.8%, outperforming SVM, Random Forest, and the baseline CNN. ROC curves, confusion matrix analysis, and performance metrics confirm the model's robustness. A lightweight, user-friendly GUI was developed to provide real-time disease prediction and recommendations for field applications. The system enables early detection, reduces misdiagnosis, and supports sustainable potato cultivation.