Classification of Geometric Shape Drawings by Using SVM and CNN Models

The landscape of education in Indonesia has been significantly shaped by globalization, with technology emerging as a pivotal influencer. This study addresses this intersection by leveraging technology to enhance teaching, focusing on preschool children’s engagement with digital tools. Researchers have developed a straightforward detection application using a convolutional neural network (CNN) as a learning tool, capitalizing on the allure of gadgets for young learners. Alternative algorithmic methods, including Support Vector Machine (SVM), were employed to assess and compare classification processes. The CNN algorithm normalizes output values and generates probabilities for each class, facilitating a comprehensive performance evaluation. This research uses 1000 images of hand-drawn simple plane figures to determine the optimal outcome, comparing accuracy values, classification reports, and confusion matrix values. The analysis reveals that the CNN model significantly outperforms the SVM model in the classification task, achieving a remarkable 100% precision, recall, and F1-score. In contrast, the SVM model attained an average precision, recall, and F1-score of 88%. This study underscores the efficacy of CNN in enhancing learning experiences for preschoolers, offering valuable insights into integrating technology into education.