Citation

Fahad, Nafiz and Hossen, Md. Jakir and Sayeed, Md. Shohel (2025) Indoor Object Detection Using YOLOv5s and Grad-CAM. In: 2025 Multimedia University Engineering Conference, MECON 2025, 21 July 2025 - 23 July 2025, Cyberjaya, Malaysia.

Text 14.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Indoor object detection is crucial for robots to navigate properly. This study integrates YOLOv5s, a lightweight and real-time object detection model, with Gradient-weighted Class Activation Mapping (Grad-CAM) for improved transparency in decision-making. A dataset comprising 10 indoor object classes was utilized, with 1,012 training, 230 validation, and 107 test images. Key optimizations such as pruning, quantization, and model width scaling were employed to enhance efficiency while maintaining detection performance. Results revealed an overall mAP@0.5 of 0.457, with "cabinet door" achieving a high mAP of 0.807, while rare classes like "pole" and "opened door" exhibited mAPs of 0.084 and 0.039, respectively. Grad-CAM visualizations highlighted critical regions influencing model predictions, enhancing interpretability. Performance metrics, including F1 scores and precision-recall curves, demonstrated the trade-offs between precision and recall. Despite challenges with underrepresented classes, this study showcases the potential of YOLOv5s and Grad-CAM for interpretable and efficient indoor object detection, addressing both computational and real-time demands in resource-constrained environments.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Indoor object detection, optimization, Yolov5
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	18 Mar 2026 08:25
Last Modified:	19 Mar 2026 02:37
URII:	http://shdl.mmu.edu.my/id/eprint/15594

Downloads

Edit (login required)