Article Per Year (5 Year)
p-Index From 2019 - 2024
0.408
P-Index
This Author published in this journals
All Journal Computatio : Journal of Computer Science and Information Systems
Devid Sumarlie
Unknown Affiliation
Computer Science & IT

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
PENGENALAN KUE TRADISIONAL INDONESIA MENGGUNAKAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK Chairisni Lubis, M.Kom.; Devid Sumarlie; Teny Handhayani
Computatio : Journal of Computer Science and Information Systems Vol. 6 No. 2 (2022): Computatio: Journal of Computer Science and Information Systems
Publisher : Faculty of Information Technology, Universitas Tarumanagara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24912/computatio.v6i2.21098

Abstract

In Indonesia, a lot of cakes are included in the category of traditional snacks. Traditional snacks are a unique culture of the archipelago that must be preserved by Indonesians. Traditional cakes are snacks that people like because they are dense and filling. Traditional cakes have a variety of textures, shapes and colors are very diverse and some are similar to each other, so it is rather difficult to identify the cake. The problem faced by buyers is that they often do not know the name of a cake because of the many types of cakes sold in the market. Technological advances have also caused many local people to use social media to take photos of food, but to recognize these cakes, there are still many people who do not really understand traditional cakes compared to modern cakes. The above problem can be solved if a system is made to recognize the image/photo of the cake and the computer can be programmed and to classify the cake into a certain category of cake by utilizing the image of the cake using the Convolutional Neural Network (CNN) algorithm. The best test results are tests that include data augmentation during training, where VGG-16 has a higher accuracy than DenseNet121 which is 80% and DenseNet121 testing which uses k-fold cross validation with an accuracy of fold 1 which is 77% and a drastic increase up to fold 5. If without using data augmentation, the best result obtained is an accuracy of 83% achieved by DenseNet121 without transfer learning, learning rate 1e-5 and batch size 16.
Co-Authors Anak Agung Gede Sugianthara Chairisni Lubis Chairisni Lubis, M.Kom.