Related Experiment Video
Updated: Jul 7, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.7K
HRBUST-LLPED: A Benchmark Dataset for Wearable Low-Light Pedestrian Detection
Tianlin Li1, Guanglu Sun1, Linsen Yu1
1School of Computer Science and Technology, Harbin University of Science and Technology, No. 52 Xuefu Road, Nangang District, Harbin 150080, China.
Micromachines
|December 23, 2023
Summary
This study introduces a new dataset and lightweight models for detecting pedestrians in low-light conditions using wearable cameras. The research enhances low-light pedestrian detection capabilities for improved safety and performance.
Area of Science:
- Computer Vision
- Robotics
- Artificial Intelligence
Background:
- Pedestrian detection in low-light conditions is a significant challenge for wearable platforms.
- Existing methods often struggle with poor illumination, limiting their real-world applicability.
- Infrared and low-light cameras offer potential solutions for enhanced detection.
Purpose of the Study:
- To introduce the HRBUST-LLPED dataset for low-light pedestrian detection.
- To develop and evaluate lightweight pedestrian detection models for wearable devices.
- To improve the accuracy and efficiency of pedestrian detection under starlight-level illumination.
Main Methods:
- Collected pedestrian data using wearable low-light cameras under starlight conditions.
- Annotated 32,148 pedestrian instances across 4269 keyframes in the HRBUST-LLPED dataset.
- Developed four lightweight pedestrian detection models based on YOLOv5 and YOLOv8, fine-tuned on the new dataset.
Main Results:
- The fine-tuned models achieved 69.90% AP@0.5:0.95.
- Inference time was recorded at 1.6 ms, demonstrating high efficiency.
- The dataset features high pedestrian density, with over seven people per image.
Conclusions:
- The developed HRBUST-LLPED dataset and models advance low-light pedestrian detection capabilities.
- This research contributes to the development of safer and more effective wearable sensing systems.
- The findings support the use of low-light cameras in wearable devices for improved pedestrian detection.
