Enhancing Potato Plant Disease Detection using CNNs with Transfer Learning: A Robust AI-Driven Approach

A crop of global importance, potatoes are particularly vulnerable to illnesses that can result in large losses for farmers. Conventional detection techniques, which depend on manual inspection, are frequently time-consuming and prone to human mistake. In order to overcome this, Transfer Learning was combined with Convolutional NeuralNetworks (CNNs) to create an improved disease detection method, utilizing pre-trained models like ResNet50 and VGG16. A collection of more than $\mathbf{3 0, 0 0 0}$ high-resolution images of potato leaves, categorized into healthy and diseased classes, was used to evaluate the effectiveness of this technique. Through model optimization and data augmentation, the methodology yielded an accuracy of about $\mathbf{9 9 \%}$, much surpassing that of conventional models such as Random Forest and Support Vector Machines, which yielded approximately 88% and 85% accuracy, respectively. Furthermore, the model’s effectiveness and scalability were demonstrated by the application of Transfer Learning, which cut training time by almost 40%. This invention provided a reliable approach for early and accurate disease management in potato cultivation, improving accuracy over traditional methods by 11%. ResNet50: Residual Network 50 layers, VGG16: Visual Geometry Group 16 layers