Apple leaf diseases critically affect fruit yield and quality. Rapid and accurate field diagnosis remains challenging due to variable lighting, complex backgrounds, and environmental noise. We propose a robust hybrid deep learning framework that effectively combines convolutional neural networks for fine‐grained local feature extraction with vision transformers for capturing global contextual dependencies and further refines them through a dual‐stage attention mechanism. It achieved state‐of‐the‐art performance, with 98.94% accuracy on the ALDD, 99.93% accuracy on the plant pathology benchmark, and a mean F1‐score of 0.992. Paired t ‐test analysis confirmed significant improvements over conventional CNN and transformer models. These results demonstrate the framework’s reliability, scalability, and potential for real‐time orchard disease monitoring, enabling precision management and sustainable apple production.
Thakur et al. (Thu,) studied this question.