Potato Leaf Disease Classification Using an Efficient Convolutional Neural Network

Early identification of potato leaf diseases plays a vital role in minimizingcrop losses and enhancing agricultural productivity. Conventional disease diagnosispractices largely depend on manual inspection by experts, which is labor-intensive,subjective, and difficult to scale in large farming environments. Although recent advancesin deep learning have enabled automated plant disease detection with high accuracy, manyexisting solutions rely on deep and computationally heavy architectures, resulting in slowtraining convergence and limited feasibility for deployment in resource-constrainedagricultural settings.This study presents a lightweight convolutional neural network specifically designed forefficient potato leaf disease classification with an emphasis on rapid training convergence.The proposed model is constructed using stacked convolutional layers integrated withbatch normalization and ReLU activation to ensure stable feature learning. To reduceparameter complexity, global average pooling is employed in place of traditional fullyconnected layers. In addition, adaptive training strategies, including the Adam optimizer,dynamic learning rate adjustment, and early stopping, are utilized to improve optimizationstability and accelerate convergence.The effectiveness of the proposed approach is validated on a three-class potato leaf imagedataset comprising Early Blight, Late Blight, and healthy leaf samples. Experimentalevaluation demonstrates that the model achieves a classification accuracy of 98% withinonly 10 training epochs, while consistently maintaining strong precision, recall, and F1-score across all categories. Performance comparison with existing deep and transferlearning–based models highlights that the proposed architecture attains comparable orsuperior results with substantially reduced training time and model complexity.The obtained results confirm that compact CNN architectures optimized for fastconvergence offer a reliable and practical solution for real-time plant disease diagnosis,particularly in low-resource and smart agriculture applications.INDEX TERMS Potato leaf disease classification, convolutional neural networks, lightweightCNN, fast convergence, plant disease detection, precision agriculture.