<![CDATA[Based on an ontological perspective, there is a gap in feature representation and in binary dysgraphia classification using ResNet18, an area that has not been explored simultaneously. Thus, our contribution is an analysis of research on dysgraphia classification using ResNet18 that employs epistemological and axiological approaches. ResNet18 was chosen as the backbone of the proposed framework because it has shortcut connections that can degrade residues into useless features. As a representation of new knowledge, ResNet18 was pre-trained on ImageNet. Classification was tested on challenging word assignments, comprising 145 dysgraphia images and 188 non-dysgraphia images. Epoch trials were conducted to find the best architecture. The results showed that ResNet18 at epoch 10 achieved the best performance in binary classification, with a recall of up to 93.55%. This indicates that ResNet18 is sensitive to recognizing dysgraphia classes. Challenges outlined in this study serve as a foundation for further research.]]>
Copyrights © 2026
