Year: 2024
Type: article
Abstract: This project addresses elevator safety by proposing a solution based on an improved lightweight YOLOv3 model. The system is trained on a custom-built dataset of electric vehicles inside elevators, achieving efficient and accurate object detection suitable for edge computing environments. It demonstrates excellent performance through transfer learning and comparative experiments. The user interface, developed with PyQt5 and Streamlit, supports image, video, and real-time detection, along with result-saving capabilities. Tests on elevator videos show outstanding accuracy and practicality, making this solution highly applicable to smart elevators and public safety management.
DOI: 10.54097/2d8v6c75
Authors: Yichen Li, Meina Zhang, Jiayu Wang, Yuxuan Liu, Yutong Xing, Xinyuan Wang
Institution: University of Science and Technology Liaoning
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4404475986