Skoči na glavni sadržaj

Engineering Review : Međunarodni časopis namijenjen publiciranju originalnih istraživanja s aspekta analize konstrukcija, materijala i novih tehnologija u području strojarstva, brodogradnje, temeljnih tehničkih znanosti, elektrotehnike, računarstva i građevinarstva, Vol. 45 No. 3, 2025.

Izvorni znanstveni članak

https://doi.org/10.30765/er.2684

Enhancing object detection in mobile augmented reality: A novel framework integrating knowledge distillation and unsupervised domain adaptation

Xiang Yun Zeng orcid id orcid.org/0009-0007-3692-0772 ; Center for Artificial Intelligence Technology, Faculty Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi 43600 Malaysia.
Siok Yee Tan ; Center for Artificial Intelligence Technology, Faculty Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi 43600 Malaysia. *
Mohammad Faidzul Nasrudin ; Center for Artificial Intelligence Technology, Faculty Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi 43600 Malaysia.
Mohammad Kamrul Hasan orcid id orcid.org/0000-0001-5511-0205 ; Faculty Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi 43600 Malaysia.

* Dopisni autor.

Puni tekst: engleski pdf 1.406 Kb

str. 94-115

preuzimanja: 50

citiraj

Sažetak

Object detection plays a crucial role in enhancing mobile augmented reality (MAR) applications, but the computational limitations of mobile devices and the dynamic real-world environments pose significant challenges. This study proposes a novel framework that integrates Knowledge Distillation (KD) and Unsupervised Domain Adaptation (UDA) to address these issues. KD transfers knowledge from a resource-heavy "teacher" model to a lightweight "student" model optimized for mobile deployment, while UDA enables the student model to adapt to real-world conditions without labeled data. Our framework uses YOLOv5 models, where the student model, YOLOv5n, learns from the teacher model, YOLOv5 small, improving precision and maintaining efficiency. Experiments on the VOC2007 and COCO datasets show that our SKD-UDA net achieves 78.2% mAP at IoU 0.5 and 50.8% mAP at IoU 0.5:0.95, outperforming the baseline YOLOv5n by 5.5% and 5.7%, respectively, without increasing the model size (1.9 MB). This approach enhances accuracy and computational efficiency, making it ideal for MAR applications. Our contributions advance object detection in MAR, improving user interaction by increasing detection accuracy, inference speed, and seamless integration of digital and physical environments.

Ključne riječi

knowledge distillation; mobile augmented reality; object detection; unsupervised domain adaptation; you only look once

Hrčak ID:

341333

URI

https://hrcak.srce.hr/341333

Datum izdavanja:

9.10.2025.

Posjeta: 150 *