Iskandar, Jonathan Steven and Kristianto, Ryan Putranda (2023) Pengenalan dan Klasifikasi Ragam Kue Indonesia menggunakan Arsitektur ResNet50V2 pada Convolutional Neural Network (CNN). Seminar Nasional Amikom Surakarta 2023, 1 (1). pp. 81-92. ISSN 3031-5581
Preview |
Text
53-File Utama Naskah-1206-1-10-20240108.pdf Download (706kB) | Preview |
Abstract
In addressing the issue of recognizing and classifying similar types of Indonesian
cakes, with a focus on cakes that exhibit high visual resemblance, this research
conducted testing of the Convolutional Neural Network (CNN) algorithm using the
ResNet50V2 architecture to identify and classify various types of Indonesian cake
images. The dataset used was obtained from Kaggle under the name "Indonesian
Cakes." This study yielded two significant findings. First, prediction results from uploaded
images were presented in the form of probabilities, and kastengel cakes were considered
valid if their prediction probabilities exceeded the threshold of 0.6. Second, the accuracy
level of the tested algorithm reached 73.19%, which can be considered a commendable
accuracy rate. These results indicate that the CNN algorithm with the ResNet50V2
architecture has the potential to be used in the recognition and classification of similar
types of Indonesian cakes, contributing to the development of an automated cake type
recognition system and aiding in maintaining quality and consistency in the food and
culinary industry. This research provides a crucial foundation for the development of
image recognition technology that can be applied within the context of Indonesian
culinary culture.
| Item Type: | Article |
|---|---|
| Uncontrolled Keywords: | CNN, ResNet50V2, Clarification, Indonesian Cake |
| Subjects: | Q Science > QA Mathematics > QA76 Computer software |
| Divisions: | Fakultas Teknik > Prodi Ilmu Informatika |
| Depositing User: | Ryan Putranda Kristianto |
| Date Deposited: | 30 Sep 2025 08:28 |
| Last Modified: | 05 Mar 2026 09:03 |
| URI: | https://repositori.ukdc.ac.id/id/eprint/1792 |
Actions (login required)
 |
View Item |