A huge part of today’s population depends on agriculture for a living, yet it is always filled with challenges accentuated by unfavorable environmental conditions and natural illnesses. This paper highlights identifying diseases in crop like potatoes. This research employs deep learning architectures like convolutional neural networks (CNNs), ResNet50, ResNet152, InceptionV3, VGG16, and VGG19 methodologies to accurately classify the potato leaf diseases such as Late Blight, Early Blight. Besides these CNN models, this paper also proposes a novel lightweight 2D CNN model for extracting the most significant features from the potato leaf images. Model interpretability is essential in deep learning since it clarifies the process and rationale behind a model’s particular predictions. This openness promotes trust, particularly in vital domains like precision agriculture and computational biology. To this end, this paper presents a comprehensive study on the use of explainable AI (XAI) model for illustrating the effectiveness of CNN architectures in classification of potato leaf diseases. Experimental results suggest that the proposed lightweight 2D CNN model attains higher accuracy compared to other CNN models.