Article Per Year (5 Year)
p-Index From 2021 - 2026
1.953
P-Index
This Author published in this journals
All Journal Jurnal Teknologi Informasi dan Ilmu Komputer Journal of Information Technology and Computer Science Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Jurnal Sistem Informasi, Teknologi Informasi, dan Edukasi Sistem Informasi (JUST-SI)
Setya Perdana, Rizal
Unknown Affiliation
Computer Science & IT Control & Systems Engineering Education Electrical & Electronics Engineering Engineering

Published : 12 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 12 Documents
Search

 1   2 
Combining Inference Results of YOLOv8 and Faster R-CNN using Weighted Boxes Fusion for Car’s Underbody Quality Inspection (Study Case On Automotive Company In Indonesia) Alfons Abilo, Nelson; Indriati; Setya Perdana, Rizal
Journal of Information Technology and Computer Science Vol. 10 No. 1: April 2025
Publisher : Faculty of Computer Science (FILKOM) Brawijaya University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25126/jitecs.2025101918

Abstract

The automotive industry in Indonesia has gradually experienced rapid development over time, leading to intense competition among automotive companies. In such a competitive environment, the quality inspection process is a critical aspect of the automotive manufacturing industry. One significant issue in the quality inspection process is human error. The Fourth Industrial Revolution offers solutions through computer vision technology to improve production quality. YOLOv8 is a real-time object detection algorithm that offers fast inference and good background reduction, but it is less effective for small and low-contrast objects. Conversely, Faster R-CNN has high confidence scores but slow inference capabilities. Company XYZ has a visual inspection system using YOLOv8, with a mAP50@95 value of 74%. However, the confidence scores produced often do not meet the company's standard threshold, which requires a confidence score above 90%. Therefore, a study was conducted to combine the inference results of YOLOv8 and Faster R-CNN using the weighted boxes fusion method to enhance the inference results of YOLOv8. The study results showed an increase in mAP when combining the inferences of the two models, compared to when each model performed inference individually – there was a 3.8% increase compared to YOLOv8's performance and a 5.5% increase compared to Faster R-CNN's performance.  
Kompresi Model YOLOv8 melalui Pruning, Sparsity, dan Knowledge Distillation untuk Deteksi Kendaraan dan Plat Nomor secara Real-Time pada Perangkat Edge Sinisuka, Gabriel Winasta; Setya Perdana, Rizal; Nurmala Sari, Irawati
Jurnal Pengembangan Teknologi dan Ilmu Komputer Vol 10 No 10 (2026): Oktober 2026
Publisher : Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Implementasi Convolutional Neural Network (CNN) pada teknologi visi komputer sering terkendala oleh tingginya kebutuhan komputasi dan konsumsi energi sehingga sulit diterapkan pada perangkat edge dengan sumber daya terbatas. Penelitian ini bertujuan mengembangkan model deteksi kendaraan dan plat nomor yang efisien namun tetap akurat dengan mengompresi arsitektur YOLOv8m. Metode yang diusulkan mengintegrasikan tiga tahap: Dynamic Sparse Training dengan regularisasi L1 pada faktor skala Batch Normalization, structured channel pruning berbasis dependency graph, serta hybrid knowledge distillation yang menggabungkan distilasi logit dan Channel-Wise Distillation (CWD) dari model teacher ke model student. Evaluasi dilakukan pada dataset gabungan MS COCO dan citra plat nomor dengan 7.344 citra validasi. Hasil eksperimen menunjukkan bahwa metode yang diusulkan mereduksi jumlah parameter sebesar 77,6% (dari 25,84 juta menjadi 5,78 juta) dan kompleksitas komputasi sebesar 74,6% (dari 78,7 menjadi 20,0 GFLOPs), serta mempercepat inferensi 1,42 kali (dari 1,7 ms menjadi 1,2 ms). Model terkompresi mempertahankan sekitar 96% akurasi model asal dengan penurunan mAP@50–95 hanya 0,022, sementara akurasi kelas plat nomor tetap terjaga pada 94%. Kombinasi pruning dan distilasi berbasis fitur terbukti efektif menghasilkan model yang ringan dan andal untuk inferensi real-time pada perangkat edge.
Co-Authors Abdul Jabbar Afrialdy, Firman Alfons Abilo, Nelson Aryo Pinandito Candra Dewi Dian Eka Ratnawati Firdaus, Nada Hamzaturrazak, Multohadi Indriati Jibril Averroes, Muhammad Lailil Muflikhah Nurmala Sari, Irawati Pandu Adikara, Putra RR. Ella Evrita Hestiandari Salsabila Sangdiva Laksono, Khansa Sinisuka, Gabriel Winasta Sulthan Rafif Susanto, Dominicus Christian Bagus Syemael Siahaan, Arion