Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal IJCCS (Indonesian Journal of Computing and Cybernetics Systems) Format : Jurnal Imiah Teknik Informatika
Delio, Ferdinand Defin
Unknown Affiliation
Computer Science & IT Control & Systems Engineering

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Perbandingan Performa Xception dan InceptionV1 untuk Pengenalan Ekspresi Wajah Delio, Ferdinand Defin; Aryani, Diah; Akbar, Habibullah; Yusuf, Mohamad; Triana, Yaya Sudarya
FORMAT Vol 15, No 1 (2026)
Publisher : Universitas Mercu Buana

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22441/format.2026.v15.i1.003

Abstract

Penelitian ini bertujuan untuk menganalisis dan membandingkan performa dua arsitektur Convolutional Neural Network (CNN) populer, yaitu Xception dan InceptionV1, dalam tugas pengenalan ekspresi wajah (Facial Expression Recognition/FER). Penelitian ini dilakukan dengan pendekatan transfer learning dan fine-tuning menggunakan dataset FER-2013 yang berisi 35.887 citra wajah grayscale berukuran 48×48 piksel yang diklasifikasikan ke dalam tujuh emosi dasar. Setiap citra diubah ukurannya menjadi 224×224 piksel, dinormalisasi, dan diproses dengan teknik augmentasi untuk meningkatkan generalisasi model terhadap variasi ekspresi wajah, pencahayaan, dan pose. Proses pelatihan dilakukan selama 30 epoch menggunakan optimizer Adam dengan learning rate 0.0001 dan batch size 64. Strategi fine-tuning dilakukan dengan membuka 30% lapisan atas model untuk mengoptimalkan bobot fitur yang telah dipelajari sebelumnya dari dataset ImageNet. Evaluasi kinerja dilakukan berdasarkan metrik akurasi, presisi, recall, F1-score, serta efisiensi komputasi yang diukur dari waktu pelatihan dan inferensi. Hasil eksperimen menunjukkan bahwa Xception mencapai akurasi validasi 70,69% dengan waktu inferensi rata-rata 20–25 ms, sedangkan InceptionV1 mencapai 65,80% dengan waktu inferensi 43–126 ms. Arsitektur Xception terbukti lebih efisien secara komputasi karena memanfaatkan depthwise separable convolution yang mengurangi jumlah parameter tanpa menurunkan akurasi. Temuan ini menunjukkan bahwa Xception lebih sesuai untuk aplikasi FER real-time dan perangkat dengan sumber daya terbatas, serta memberikan dasar yang kuat bagi penelitian lanjutan dalam pengembangan sistem pengenalan ekspresi wajah berbasis video dan lingkungan dunia nyata.
Comparative Performance Analysis of Xception and ResNet50 Architectures for Facial Expression Recognition Using the FER-2013 Dataset Aryani, Diah; Akbar, Habibullah; Delio, Ferdinand Defin
IJCCS (Indonesian Journal of Computing and Cybernetics Systems) Vol 20, No 3 (2026): July
Publisher : IndoCEISS in colaboration with Universitas Gadjah Mada, Indonesia.

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22146/ijccs.117591

Abstract

This study aims to evaluate and compare the performance and computational efficiency of the Xception and ResNet50 architectures in facial expression classification tasks. Facial Expression Recognition (FER) plays an important role in the development of intelligent systems capable of interpreting human emotions. This technology has various applications, including adaptive learning systems, emotion-aware customer service, and mental health support systems. This research compares two widely used Convolutional Neural Network (CNN) architectures, Xception and ResNet50, for facial expression classification using the FER-2013 dataset. The dataset contains 35,887 grayscale facial images with a resolution of 48×48 pixels categorized into seven basic emotions. All images were resized to 224×224 pixels and converted into RGB format to match the input requirements of pretrained ImageNet models.Both architectures were trained using a transfer learning strategy with selective fine-tuning on specific layers. Data augmentation techniques were applied to increase dataset variability and reduce overfitting. Model performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrix metrics. The results show that Xception architecture outperforms ResNet50, achieving a validation accuracy of 70.69% and a weighted F1-score of 0.71. These findings demonstrate that appropriate architecture selection and structured training strategies can significantly improve FER performance in practical intelligent systems.
Co-Authors Akbar, Habibullah Diah Aryani, Diah Mohamad Yusuf Yaya Sudarya Triana