Implementasi Media Pembelajaran Interaktif Pengenalan Sayuran Untuk Anak Usia Dini Menggunakan Convolutional Neural Network

Salsabila, Akila Nor (2026) Implementasi Media Pembelajaran Interaktif Pengenalan Sayuran Untuk Anak Usia Dini Menggunakan Convolutional Neural Network. Undergraduate thesis, Politeknik Negeri Jember.

	Text (Abstract) Abstract.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (211kB)
	Text (Bab 1 Pendahuluan) Pendahuluan.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (210kB)
	Text (Daftar Pustaka) Daftar.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (204kB)
	Text (Laporan Lengkap) FIX.pdf - Submitted Version Restricted to Registered users only Download (7MB) | Request a copy

Abstract

This study aims to implement interactive learning media for introducing vegetables to early childhood based on a website using Convolutional Neural Network (CNN). The media is equipped with image classification features, bilingual audio (Indonesian and English), and educational games. The architectures compared are ResNet50 and EfficientNet-B0 with transfer learning on 30 vegetable classes. The training results show that ResNet50 obtained a training accuracy of 95.47% and a validation accuracy of 87.41%, while EfficientNet-B0 obtained a training accuracy of 87.50% and a validation accuracy of 87.22%. Based on these results, optimization was performed on EfficientNet-B0 because it had more stable performance. Optimization was performed through intensive augmentation (rotation, shift, zoom, brightness, channel shift, and horizontal flip), fine-tuning of the last 40 layers, addition of a dense layer of 128 neurons, and a learning rate of 5e-5 with ReduceLROnPlateau. The optimization results increased the validation accuracy to 93.33% and the testing accuracy based on the confusion matrix to 96.48%. Black box testing showed that all features worked well. Evaluation using pre-test and post-test showed an increase in learning outcomes, with the average score rising from 38.43% to 82.41%. Furthermore, UAT obtained a score of 94.67% with a very good category. Thus, this media is declared effective and suitable for use as a learning tool for vegetable recognition for early childhood.

Actions (login required)

View Item