Article Per Year (5 Year)
p-Index From 2020 - 2025
0.23
P-Index
This Author published in this journals
All Journal Jurnal Ilmiah Teknik Elektro Komputer dan Informatika (JITEKI)
Iqbal Kurniawan Perdana
Telkom University
Computer Science & IT Electrical & Electronics Engineering

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

Found 1 Documents
Search

 1 
Experimenting with the Hyperparameter of Six Models for Glaucoma Classification Muhammad Ilham; Angga Prihantoro; Iqbal Kurniawan Perdana; Rita Magdalena; Sofia Saidah
Jurnal Ilmiah Teknik Elektro Komputer dan Informatika Vol 9, No 3 (2023): September
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26555/jiteki.v9i3.26331

Abstract

Glaucoma, being one of the leading causes of blindness worldwide, often presents without noticeable symptoms, making early detection crucial for effective treatment. Numerous studies have been conducted to develop glaucoma detection systems. In this particular study, a glaucoma detection system using the CNN method was developed. The models employed in this study include AlexNet, Custom Layer, MobileNetV2, EfficientNetV1, InceptionV3, and VGG19. For training, an augmented RIM-ONE DL dataset was utilized. Hyperparameter experiments were conducted to determine the most optimal parameters for each model, specifically testing batch size, learning rate, and optimizer. The hyperparameter optimization process yielded the optimal parameters for each model. However, it is important to note that the MobileNetV2, InceptionV1, and VGG19 models exhibited signs of overfitting in the training graph results. Among the models, the custom layer model achieved the highest accuracy of 93%, while InceptionV3 attained the lowest accuracy at 83.5%. Testing of the models was performed using data from Cicendo Eye Hospital and the RIM-ONE DL testing dataset. Based on the testing results, it was found that InceptionV3 outperformed the other models in predicting images accurately. Therefore, the study concluded that high accuracy in training does not necessarily indicate superior performance in testing, particularly when limited variation exists in the training dataset.
Co-Authors Angga Prihantoro Muhammad Ilham Rita Magdalena SOFIA SAIDAH