Customized CNN Model for Multiple Illness Identification in Rice and Maize

Abstract

Crop diseases imperil global food security and economies, demanding early detection and effective management. Convolutional Neural Networks (CNNs), particularly in rice and maize leaf disease classification, have gained traction due to their automatic feature extraction capabilities. CNN models eliminate manual feature extraction, enabling precise disease diagnosis based on learned features. Researchers have rapidly advanced these models, achieving promising results. Leaf disease characteristics like color changes, texture variations, and lesion appearance have been identified as useful for automated diagnosis using machine learning. Developing CNN models involves crucial stages: dataset preparation, architecture selection, hyperparameter tuning, and model training and evaluation. Diverse and accurately annotated datasets are pivotal, and appropriate CNN architecture selection, such as ResNet101 and XceptionNet, ensures optimal performance. These architectures' pre-training on vast image datasets enhances feature extraction. Hyperparameter tuning fine-tunes the model, and training and evaluation gauge its precision. CNN models hold potential to enhance rice and maize productivity and global food security by effectively detecting and managing diseases.