Review: Advanced Deep Learning Models for Real-World Plant Disease Detection and Classification

The integration of deep learning into plant pathology has transformed the detection and classification of plant diseases, addressing critical challenges in global food security. This systematic review evaluates the progression and efficacy of deep learning architectures for real-world plant disease identification, analyzing 127 studies from 2015 to 2024. We assess foundational Convolutional Neural Networks (CNNs), advanced models like DenseNet and EfficientNet, and emerging Vision Transformers, emphasizing their performance under field conditions. While accuracies exceed 99% in controlled settings, variability in environmental factors poses persistent hurdles. Through comparative tables, we detail architectural strengths, dataset characteristics, augmentation strategies, and performance metrics, incorporating a novel multi- model framework as a case study. This work delineates research gaps and future trajectories, offering a technical foundation for scalable, robust solutions in agricultural diagnostics.