<![CDATA[Corn is one of the most important agricultural commodities in the Toba region of North Sumatra. However, its productivity is often reduced due to foliar diseases that appear prior to harvest. The three most commonly observed leaf diseases include leaf spot, blight, and rust. To support early detection efforts among local farmers, this study proposes an image-based classification system employing the Convolutional Neural Network (CNN) algorithm and the DenseNet121 model as a transfer learning approach. The primary objective of this research is to automatically identify the type of disease affecting corn leaves using image data, thereby enabling farmers to promptly implement appropriate countermeasures. A series of experiments were conducted to evaluate various model configurations, including different activation functions (ReLU and Tanh), adjustments to learning rates, and the tuning of other hyperparameters such as optimizers and preprocessing methods (normalization, rotation augmentation, zooming, and contrast adjustments). The results demonstrate that DenseNet121, when trained with an optimal learning rate of 0.001, achieved the highest accuracy of 97%, outperforming the custom-built CNN model which attained an accuracy of 95%. The combination of effective preprocessing techniques and hyperparameter tuning significantly contributed to the improved performance of the models. This study highlights the potential of image-based plant disease detection technologies in agriculture, particularly in aiding real-time decision-making, enhancing land management efficiency, and supporting increased corn yield.]]>
