<![CDATA[Rice foliar diseases pose a major threat to global food security by reducing yield and grain quality, motivating the need for scalable, objective, and automated diagnosis solutions. This study investigates the impact of Contrast Limited Adaptive Histogram Equalization (CLAHE) and transfer learning on the classification of three common rice leaf diseases—Bacterial Blight, Brown Spot, and Leaf Smut—from RGB leaf images. Using a dataset of 2,342 images split into training, validation, and test sets (80:10:10), we design a controlled experimental pipeline comprising four scenarios: with/without CLAHE, and with/without transfer learning. CLAHE is applied as a preprocessing step to enhance local contrast and lesion visibility under heterogeneous illumination and cluttered backgrounds, while transfer learning leverages ImageNet-pretrained convolutional neural networks fine-tuned for rice disease recognition. Models are trained and evaluated using accuracy, macro F1, and weighted F1 on a held-out test set. The combined CLAHE + transfer learning configuration achieves the best performance, with an overall accuracy of 0.94 and macro and weighted F1-scores of 0.94, substantially outperforming non-enhanced and non-transferred baselines. Qualitative analysis indicates improved separability between visually similar classes, particularly Brown Spot and Leaf Smut, under challenging imaging conditions. These findings underscore the effectiveness of integrating contrast enhancement with transfer learning for robust, field-oriented rice disease classification and highlight a practical pathway toward reliable image-based decision support in precision agriculture.]]>
Copyrights © 2026
