Plant leaf diseases and variations in growth stages have a direct impact on crop yield and overall food production, making early and accurate identification essential for effective plant management. In many cases, farmers still depend on visual inspection to recognize these conditions, but this method is often slow, subjective, and inconsistent, especially when different diseases exhibit similar symptoms in natural field environments. Image-based analysis provides a more practical alternative; however, traditional approaches usually rely on hand-crafted features, which often fail to perform well under variations in lighting, background, and leaf appearance. Deep learning has demonstrated strong potential for overcoming these limitations by automatically learning meaningful features from images, particularly through convolutional neural networks that can capture color, texture, and shape patterns. Still, the performance of a single model may vary across different disease categories and growth stages because of differences in visual complexity. Therefore, this study proposes a deep learning approach for plant leaf disease and growth stage classification using real-field images, with particular emphasis on image quality, model training, and performance evaluation, in order to develop a reliable and practical solution for automated plant health monitoring in real agricultural settings.
Fahim et al. (Tue,) studied this question.