Claim Missing Document
Check
Articles

Found 3 Documents
Search
Journal : International Journal of Artificial Intelligence and Science

Vehicle Detection on The Traffic Using Detection Transformer (DETR) Algorithm Khoiriyah, Rofiatul; Hendrawan, Aria
International Journal of Artificial Intelligence and Science Vol. 1 No. 1 (2024): September
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63158/IJAIS.v1.i1.4

Abstract

Object detection is a computer vision technique aimed at detecting and identifying objects in images or videos. In recent years, with advancements in Machine Learning and Deep Learning, object detection has made significant progress in various fields such as healthcare, security, and transportation. The DETR algorithm is a novel approach in object detection that combines transformer architecture with attention techniques to address object detection challenges. This research applies the DETR algorithm with ResNet backbone for vehicle detection on the roads, involving 6 object classes: Car, Truck, Bus, Motorcycle, Pickup Car, and Truck Box. Four training experiments were conducted: DETR-ResNet50, DETR-ResNet101, DETR-DC5-ResNet50, and DETR-DC5-ResNet101. The implementation results show that DETR-DC5 improves the accuracy of vehicle detection. DETR-DC5 with ResNet-101 achieved the highest score for AP50, which is 0.957. However, it should be noted that DETR-DC5 with ResNet-50 managed to maintain overall AP stability, with a lower parameter of 35.5. The model's outcomes in this study can be effectively applied for vehicle detection on the roads.
Traffic Vehicle Detection Using Faster R-CNN: A Comparative Analysis of Backbone Architectures Hakim, Luqman; Hendrawan, Aria; Khoiriyah, Rofiatul
International Journal of Artificial Intelligence and Science Vol. 1 No. 1 (2024): September
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63158/IJAIS.v1.i1.5

Abstract

Object detection is a crucial task in computer vision, where advanced deep learning models have shown significant improvements over traditional methods. In this study, the Faster R-CNN algorithm is applied to a traffic dataset containing six vehicle categories: Bus, Car, Motorcycle, Pick Up Car, Truck, and Truck Box. The novelty of the research lies in the comparison of four backbone architectures ResNet50, ResNet50V2, MobileNetV3 Large, and MobileNetV3 Large 320 evaluated for their performance in vehicle detection at IoU thresholds of 0.5 and 0.75. The results reveal that ResNet50 provided the best overall performance, achieving mAP scores of 0.966 at IoU 0.5 and 0.887 at IoU 0.75, offering a balanced trade-off between precision and recall. ResNet50V2 and MobileNetV3 Large also performed well, with mAP scores of 0.945 and 0.870 for ResNet50V2, and 0.969 and 0.843 for MobileNetV3 Large, respectively. However, MobileNetV3 Large 320 showed the lowest detection performance, with mAP scores of 0.857 at IoU 0.5 and 0.551 at IoU 0.75. These findings provide useful insights into the suitability of different architectures for vehicle detection tasks, particularly in traffic surveillance applications.
A Hybrid Image Processing Approach for Real-Time Face Recognition in Attendance Monitoring Agho, Davit Cany; Hendrawan, Aria
International Journal of Artificial Intelligence and Science Vol. 2 No. 1 (2025): March
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63158/IJAIS.v2.i1.17

Abstract

In the era of digital transformation, institutions are increasingly adopting automation to enhance administrative efficiency, particularly in human resource management. At Tanggirejo Village Hall, a critically low employee attendance rate of 46.45% in January 2024 exposed the limitations of manual attendance systems, which are prone to errors and manipulation. This study proposes a face recognition-based attendance system utilizing OpenCV’s Haar Cascade Classifier for face detection and the Local Binary Pattern Histogram (LBPH) for face recognition. A total of 500 grayscale facial images from 10 employees were collected and processed to train and test the system. Evaluation using a Confusion Matrix revealed an accuracy of 72%, precision of 93%, and recall of 75%. While a 27% error rate was observed—primarily due to lighting inconsistencies and limited training data—the system performed reliably in real-time scenarios. The integration of these lightweight algorithms allows for fast and accurate identification, suitable for resource-constrained environments. This solution not only addresses the local attendance challenges but also presents a scalable, automated model that can be adopted by similar institutions seeking to improve productivity and operational transparency through real-time employee monitoring.