Multi-Crop Disease Prediction Using Attention-Augmented Deep Learning: A DenseNet121-CBAM Model

Plant health is essential for maintaining agricultural output and economic stability, as they are essential for supporting life and global economies. Diseases brought on by pathogens, such as insects, fungus, and bacteria, can severely impact plant health, emphasizing the significance of early plant disease classification and identification. In order to detect and categorize plant leaf diseases, this work suggests a deep learning method that utilizes the DenseNet121 architecture and is improved by a CBAM. CBAM is a versatile, lightweight component capable of being trained in an end-to-end manner alongside foundational CNNs and can be seamlessly incorporated into any CNN architecture with little additional burden. Experimental results demonstrate that DenseNet121+CBAM achieves a remarkable accuracy of 98.88% for a four-crop (tomato, potato, corn, pepper) classification task, significantly outperforming MobileNetV1+CBAM. Additional experiments on a three-crop subset (tomato, potato, and pepper) reaffirmed the superiority of DenseNet121+CBAM with an accuracy of 98.23%, marginally exceeding MobileNetV2+CBAM and MobileNetV1+CBAM. These findings highlight the robustness and scalability of DenseNet121+CBAM, making it highly suitable for large datasets and multi-crop classification tasks. This outstanding performance emphasizes its potential for broader agricultural applications, enabling efficient and accurate disease detection to support sustainable farming practices.