A Comparative Study of Object Detection Algorithms for Security Applications

Keywords:
Artificial Intelligence, Computer Vision, Classification, Object Detection, Security Applications

Abstract:
Currently, there are many different computer vision models available, and each has its unique characteristics. Selecting the most suitable and, importantly, well-performing model can be challenging for companies and researchers who plan to use artificial intelligence to solve their problems. This study aims to evaluate the performance of four prominent computer vision models: YOLOv5, Faster R-CNN, SSD 300, and RetinaNet. The models were assessed on their ability to detect and classify weapons in images. The primary metrics used for evaluating their performance are mAP@50 and mAP@50-95. The dataset used for testing these models is taken from the well-known dataset platform Kaggle and consists of images of various types of weapons sorted by class. This circumstance also makes it possible to associate this research with the field of security and its automation. Experimental results identified YOLOv5 as the best-performing model among the four. The overall performance was constrained by the dataset’s limited size and image quality, with the highest mAP@50 reaching 0.8. The findings of this study offer practical insights for companies seeking effective computer vision solutions, as well as for researchers examining the development and comparative performance of object detection models.

CITATION:

IEEE format

Copy to clipboard

R. Kriuchkov, T. Bezdan, “A Comparative Study of Object Detection Algorithms for Security Applications,” in Sinteza 2025 - International Scientific Conference on Information Technology, Computer Science, and Data Science, Belgrade, Singidunum University, Serbia, 2025, pp. 524-530. doi:10.15308/Sinteza-2025-524-530

APA format

Copy to clipboard

Kriuchkov, R., Bezdan, T. (2025). A Comparative Study of Object Detection Algorithms for Security Applications. Paper presented at Sinteza 2025 - International Scientific Conference on Information Technology, Computer Science, and Data Science. doi:10.15308/Sinteza-2025-524-530

BibTeX format
Download

RefWorks Tagged format
Download