A Deep Learning Approach for Classifying Severity of Dysgraphia

The ability to accurately detect and assess the severity of dysgraphia is pivotal for providing timely intervention and support to affected individuals. This paper explores the application of Deep learning model to evaluate the severity of dysgraphia by analyzing handwriting samples. VGG16 model, a trained CNN renowned for its exceptional ability to recognize images is employed. The objective is to classify handwriting samples into 'Low Potential Dysgraphia' and 'Potential Dysgraphia'. The Dysgraphia image datasets is preprocessed to align with the VGG16 model’s input requirements to ensure reliability by adjusting each images resolution to 224 x 224 pixels and using a batch size of 32. The model is trained over 10 epochs for optimal learning performance. The VGG16 model is effectively assessed based on the performance metrics viz; accuracy, precision, loss, recall, F1 score. VGG16 significantly achieves a test accuracy rate of 92%, showcasing its proficiency in classifying handwriting associated with dysgraphia.