Aisy, Rifa Amalia Rahadatul (2026) IMPLEMENTASI ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN) UNTUK PENGENALAN BENTUK GEOMETRIS PADA ANAK USIA DINI (STUDI KASUS : KELOMPOK BERMAIN BENTANG SANGKANMULYA). S1 / D3 thesis, Universitas Kuningan.
ABSTRAK.pdf
Download (854kB)
BAB I.pdf
Download (792kB)
BAB II.pdf
Restricted to Repository staff only
Download (707kB) | Request a copy
BAB III.pdf
Restricted to Repository staff only
Download (1MB) | Request a copy
BAB IV.pdf
Restricted to Repository staff only
Download (1MB) | Request a copy
BAB V.pdf
Download (341kB)
DAFTAR PUSTAKA.pdf
Download (289kB)
LAMPIRAN.pdf
Restricted to Repository staff only
Download (2MB) | Request a copy
Abstract
Pendidikan anak usia dini (PAUD) krusial untuk pengembangan kognitif, termasuk pengenalan bentuk geomertis 2D yang membantu pemahaman pola spasial. Namun, metode konvensional seperti origami kurang menarik, diperparah keterbatasan media di KB Bentang Sangkanmulya, sehingga kurang efektif. Untuk mengatasi masalah tersebut, dikembangkan aplikasi android berbasis Convolutional Neural Network (CNN) dengan arsitektur Xception untuk mengenali tujuh bentuk lingkaran, segitiga, bujur sangkar, persegi panjang, trapesium, segilima, dan belah ketupat. Metodologi meliputi observasi, wawancara, tinjauan pustaka, dan pengembangan via Prototyping. Aplikasi menggunakan Tensorflow Lite untuk inferensi real-time, dilengkapi audio dan permainan edukasi. Dataset yang digunakan 7.000 citra (80% pelatihan, 10% validasi dan 10% test) dengan augmentasi. Pengujian melalui Confusion Matrix menunjukkan akurasi 94,32%, F1-Score tinggi per bentuk, serta deteksi akurat pada jarak 10-50 cm.
Early childhood education (PAUD) is crucial for cognitive development, including the introduction of 2D geometric shapes that help children understand spatial patterns. However, conventional methods like origami are less engaging, exacerbated by limited media variety at Kelompok Bermain Bentang Sangkanmulya, making them less effective. To address this, an Android application based on Convolutional Neural Network (CNN) with Xception architecture was developed to recognize seven shapes: circle, triangle, square, rectangle, trapezoid, pentagon, and rhombus. The methodology includes observation, interviews, literature review, and development via Prototyping. The app uses Tensorflow Lite for real-time inference, equipped with audio and educational games. The dataset consists of 7000 images (80% training, 10% validation, and 10% test) with augmentation. Testing through Confusion Matrix shows 94.32% accuracy, high F1-Score per shape, and accurate detection at distances of 10-50 cm.
| Item Type: | Thesis (S1 / D3) |
|---|---|
| Uncontrolled Keywords: | Convolutional Neural Network, Bentuk Geomertis, Anak Usia Dini, Xception, Convolutional Neural Network, Geometric Shape Recognition, Early Childhood, Xception. |
| Subjects: | T Technology > T Technology (General) T Technology > T Technology (General) > T201 Patents. Trademarks |
| Divisions: | Fakultas Ilmu Komputer > S1 Teknik Informatika |
| Depositing User: | S.Kom Rifa Amalia Rahadatul Aisy |
| Date Deposited: | 22 May 2026 01:58 |
| Last Modified: | 22 May 2026 01:58 |
| URI: | https://rama.uniku.ac.id/id/eprint/5159 |

