In this work, a Saliency-Guided ResNet34 model is introduced for the classification of corn leaves as healthy and infected. The dataset contains 4226 images collected from Kaggle and was divided into training, validation, and test subsets. Preprocessing steps such as resizing, normalization, and data augmentation were used to improve generalization, and transfer learning using ImageNet weights was used to take advantage of ResNet34’s powerful feature extraction ability. The overall accuracy of the model was 91% after 15 epochs, with precision, recall, and F1 scores exceeding 0.90 for both classes. Confusion matrix analysis showed a great balance between sensitivity and specificity with fewer misclassifications. Furthermore, the saliency mapping was used to provide interpretability in terms of localizing symptomatic leaf regions that were consistent with the visual pathology and ensuring transparency to decision-making. These results validate that deep learning models, particularly ResNet34 with saliency guidance, can be an important part of early disease detection and precision agriculture. Additional research on multi-class classification, integration with IoT-based systems, and real-world deployment using mobile and drone-assisted platforms will be identified for further development of sustainable and intelligent crop health monitoring.