<p>Timely detection of plant diseases is vital for protecting crop yields, ensuring food security, and maintaining ecological stability. While deep learning has shown significant promise in automating plant disease detection, many existing methods overlook the considerable variation in image datasets arising from diverse lighting conditions, camera settings, and environmental factors. The present study addresses this challenge by incorporating global color constancy algorithms such as Color Opponency Space (HSV), Max RGB, and Shades of Gray into the preprocessing pipeline to enhance image uniformity prior to classification. Four convolutional neural network architectures (VGG19, Inception&#xa0;v3, ResNet50, and DenseNet121) were evaluated on an apple leaf disease dataset comprising four classes. Results indicate that ResNet50, when combined with the HSV-based color constancy method, outperformed other configurations with the highest classification accuracy and lowest validation loss. The findings validate the effectiveness of integrating color constancy techniques with deep learning for robust plant disease classification.</p>