Novel Plant Leaf Disease Detection Approach using Hybrid Deep Learning Strategy

The early detection and identification of plant diseases are pivotal for precision agriculture, aiming to mitigate crop losses and optimize yields. This study presents a novel approach to plant disease identification, leveraging Deep Learning (DL) techniques with a focus on transfer learning. Traditional methods often rely on complex image enhancement, disease region segmentation, and feature extraction, whereas our approach employs Convolutional Neural Networks (CNNs) for efficient and accurate disease detection. To enhance accessibility and diagnostic capabilities, we implement Hybrid DL Strategy, which consists of EfficientNetB0 and MobileNetV2 model, tailored for lightweight applications suitable for smartphone implementation. Efficient NetB0 model, which considers depth, width, and resolution during convolution, enhancing the model's capacity to capture intricate features critical for accurate disease diagnosis. The overall accuracy of proposed method is 98.44%, surpassing CNN, ResNet, and AlexNet. This high accuracy underscores the effectiveness of the proposed model in correctly classifying both diseased and healthy plant leaves.