PlantNet-Lite: A Lightweight CNN Approach for Plant Disease Recognition

Plant diseases like rust and powdery mildew pose serious risks to agriculture worldwide and can have a devastating effect on harvests if they are not identified in time. While powdery mildew damages plants with white or gray powder-like spots, rust causes pustules on leaves and stems that hinder photosynthetic activity. These illnesses have the potential to spread quickly, resulting in large crop losses, upsetting food supply systems, and hurting farmers' bottom lines. Recent developments in deep learning have made it possible to automatically diagnose diseases using pretrained models such as VGG19 and MobileNetV2, which are crucial for early identification. Nevertheless, these models frequently necessitate significant computational resources, rendering them unsuitable for implementation in settings with limited resources. A lightweight convolutional neural network (LCNN) for effectively identifying plant diseases from leaf photos is presented in this paper named PlantNet-Lite. PlantNet-Lite preserves competitive accuracy while lowering computing complexity. In testing, it outperformed more complicated models in terms of efficiency and resource utilization, achieving 96.88% test accuracy. PlantNet-Lite is appropriate for use on mobile devices and in settings with constrained processing resources due to its lower size and quicker inference time. This study opens up possibilities for more improvement and scalability in practical agricultural applications and shows how LCNNs can offer a balanced solution for plant disease diagnosis by combining high accuracy with useful efficiency.