<scp>PLDC</scp> ‐Net: Potato Leaf Disease Classification Network Based on an Efficient Convolutional Neural Network

ABSTRACT Agriculture plays a key role in the economic growth of a country, and it is a primary source of food production for both humans and animals. Plant diseases, on the other hand, are a major issue that have a significant impact on agriculture. The yearly loss of productivity in agriculture due to these diseases is ~25%. Detecting plant diseases quickly and accurately is crucial for improving crop yields. In this study, a novel approach named PLDC‐Net was developed based on convolutional neural networks (CNNs) with transfer learning to detect multiple diseases of potato crops. The proposed study focused on data balancing, a key aspect of deep learning (DL) useful for improved accuracy and generalization, as usually imbalanced data hinders model generalization. To train the suggested approach, a large set of photos was collected from various online resources, and the balance between the target classes was ensured. Pre‐processing and data augmentation techniques were applied before training the model to enhance generalization and reduce overfitting. A pre‐trained model, EfficientNet‐B1, was used as the backbone of the model to extract the high‐level features of leaf images and was fine‐tuned with two dense layers, followed by SVM as the output layer for disease type identification. The model was evaluated on another set of unseen images with 98.39% average accuracy. The proposed network may provide a reliable and effective way of identifying potato diseases, hence ensuring food security and reducing agricultural financial losses. This work highlights how CNN algorithms can be fine‐tuned and employed to categorize potato illnesses, thereby enabling automatic and effective control of diseases in potato agricultural production.