Corn leaf diseases can reduce crop productivity by disrupting photosynthesis and plant growth. Manual identification in large-scale fields remains limited due to its dependence on observer expertise, visual similarity among disease symptoms, and variations in field conditions. This study proposes CornLeafNet, a Custom Convolutional Neural Network model for disease-area-based corn leaf disease classification. The dataset consists of XML-annotated corn leaf images, from which disease-affected regions were extracted through annotation parsing and bounding box-based cropping to focus the model on symptomatic leaf areas. CornLeafNet was developed to classify three disease categories: Grey Leaf Spot, Corn Rust, and Leaf Blight. The model achieved a validation accuracy of 97.66% and a testing accuracy of 98.60%, with precision, recall, and F1-score values of 0.9860, respectively. The best-performing model was converted into ONNX format and deployed in a web-based prototype for image- and video-based classification. The testing results showed that all core system functions operated as expected, indicating that CornLeafNet has potential as an automatic and practical support model for early corn leaf disease identification.