AI in Agriculture: A Review of Deep Learning-Based Crop Disease Detection

Abstract: International food security is largely dependent on farming, but crop diseases continue to threaten crop quality and yield. To minimize losses and maintain sustainable agriculture, plant diseases must be accurately and promptly diagnosed. Automation is very desirable because the traditional methods of visual inspection and laboratory analysis are frequently laborious, subjective, and unavailable in remote locations. Crop disease detection has been transformed by recent advances in deep learning (DL), which enable models to automatically extract discriminative features from plant photos without the need for human assistance. From manual scouting and preliminary image processing to conventional machine learning and more recent state-of-the-art deep architectures, this review tracks the development of disease detection techniques. Across popular crops like rice, maize, and tomatoes, pivotal techniques like convolutional neural networks, transfer learning, ensemble methods, and vision transformers are critically reviewed and compared. Examined are real-world uses like drone imaging, precision agriculture systems, mobile applications, and IoT-based monitoring. Along with fascinating potential directions like multimodal learning, cloud–edge AI fusion, and farmer-centric design, challenges like sparse datasets, environmental heterogeneity, computational cost, and unexplainability are discussed. This review provides a comprehensive picture of creating reliable, field-deployable crop disease detection systems by synthesizing improvements and shortcomings.

Keywords: Agriculture; Deep Learning; Crop Disease Detection; Precision Farming; Convolutional Neural Networks; Vision Transformers.