<![CDATA[Tea leaf diseases negatively affect both crop quality and productivity, highlighting the importance of early and accurate diagnosis. This study evaluates ResNet50 and ResNet101 as pretrained feature extractors combined with a Support Vector Machine (SVM) to determine whether a deeper residual architecture can better distinguish visually similar tea leaf disease classes. Images were obtained from the Kaggle “Identifying Disease in Tea Leaves” dataset, comprising eight categories: Anthracnose, algal leaf, bird eye spot, brown blight, gray light, healthy, red leaf spot, and white spot. Data augmentation increased each class to 800 images, producing 6,400 images that were divided into training and testing sets using an 80:20 ratio. Images were resized to 128 × 128 pixels for ResNet50 and 224 × 224 pixels for ResNet101. Feature vectors were generated by flattening convolutional feature maps from the ImageNet-pretrained networks without additional pooling. The extracted features were standardized and classified using an RBF-kernel SVM with C = 1.0 and gamma set to “scale.” ResNet101–SVM achieved 97.97% test accuracy, with precision, recall, and F1-score of 98%. In comparison, ResNet50–SVM obtained 89.15% accuracy, 90% precision, 89% recall, and 89% F1-score. These results indicate that ResNet101 captures finer texture and lesion-boundary information, enabling more reliable discrimination among similar disease classes. Therefore, deeper residual feature extraction provides a promising foundation for automated tea leaf disease diagnosis.]]>
