Related Experiment Video
Updated: Mar 28, 2026
05:30
Large Scale Energy Efficient Sensor Network Routing Using a Quantum Processor Unit
Published on: September 8, 2023
1.2K
An adaptive multi-scale lightweight network for long-distance small traffic sign detection
Ruishi Liang1,2, Wenjie Qu3, Shuaibing Li4
1School of Computer Science, University of Electronic Science and Technology of China, Zhongshan Institute, Zhongshan, 528402, China. liangruishi@foxmail.com.
Scientific Reports
|March 14, 2026
Summary
We developed YOLO-AML, an efficient system for detecting small traffic signs in autonomous driving. It enhances accuracy and speed while reducing computational load, improving road safety.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Robotics
Background:
- Autonomous driving systems require precise detection of small, distant traffic signs.
- Existing algorithms struggle with subtle features, background clutter, and real-time processing demands.
Purpose of the Study:
- To develop an efficient and accurate object detection model for small traffic signs in autonomous driving.
- To address limitations in current algorithms concerning feature preservation, background noise, and computational complexity.
Main Methods:
- Introduced YOLO-AML, utilizing parameter-free spatial transformations and low-channel convolutions.
- Proposed the Normalization-based Attention with sigmoid and tanh (NAST) module for attention regulation and noise suppression.
- Integrated the C2PSA-LSKA (CLSKA) module to enhance receptive fields and reduce parameters.
- Employed a Normalized Wasserstein Distance (NWD) loss function to handle small object detection challenges.
Main Results:
- Achieved a 17% reduction in parameters and 16.8% decrease in computational complexity.
- Reached a detection speed of 72.2 FPS with a 2.0% improvement in detection accuracy.
- Grad-CAM visualizations confirmed enhanced feature discriminability and robustness against background interference.
Conclusions:
- YOLO-AML significantly improves small traffic sign detection performance in complex driving scenarios.
- The model offers a computationally lightweight and efficient solution for autonomous driving perception systems.