Channel Attention-Guided Lightweight CNN Meets Extreme Learning Machine: Multi-Class Crop Leaf Disease Classification with Explainable AI

Crop leaf diseases pose a significant challenge to global agriculture, leading to substantial economic losses and reduced yields. These ailments, caused by various pathogens, manifest as symptoms such as leaf discoloration and impaired photosynthesis, severely impacting crop health and productivity. Existing classification methods face challenges related to accuracy, computational efficiency, and interpretability. This study introduces an innovative solution that combines a meticulously designed lightweight convolutional neural network (CNN) architecture with a channel attention (CA) mechanism for extracting crucial features and leverages the discriminative power of an extreme learning machine (ELM) for classification. The lightweight CNN developed demonstrates significantly reduced training times. Additionally, the CA mechanism within the model effectively prioritizes channel-wise critical features while eliminating irrelevant and redundant ones. An empirical evaluation is conducted using a publicly available dataset and the proposed model achieves a flawless 100% accuracy in distinguishing between 39 distinct crop leaf diseases. Furthermore, the study employs gradient class activation maps (Grad-CAM) visualization to delve into the model’s decision-making process, shedding light on influential factors and regions. It holds the promise of transforming crop health and agricultural yields, instilling hope for farmers and researchers dedicated to achieving agricultural prosperity and sustainability.