Enchancement of convolutional neural network through augmentation in zoo animal image recognition for toddlers: a hybrid approach using teachable machine.

ABSTRACT: Young children are exposed to technology more and more in the modern digital age, which has a big impact on how they learn. As mobile devices and interactive applications become more common, there is an increasing demand for educational resources that are especially tailored to the developmental needs of young children. Artificial intelligence (AI) and early childhood education have a lot to offer when it comes to utilizing cutting-edge technology to improve learning in a fun and interactive way. In order to improve Convolutional Neural Networks (CNNs) ability to identify photos of zoo animals for younger learners, this thesis investigates the integration of image augmentation techniques. The work uses augmentation techniques to increase the model’s generalization across various image kinds and situations and enrich the training dataset using the Teachable Machine platform. A comparison of the enhanced prototype and the current system showed notable advancements. Through three testing, the augmented prototype’s accuracy range from 80% to 83%, proving the usefulness of augmentation techniques in improving model performance-particularly in identifying animals from different perspectives.

Filipiniana

5