Related Experiment Video
Updated: May 12, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
432
A complex roadside object detection model based on multi-scale feature pyramid network.
Zhihao Zheng1, Jianguang Zhao1, Jingjing Fan2
1College of Information Engineering, Hebei University of Architecture, 13 Chaoyang West Street, Zhangjiakou, 075000, People's Republic of China.
Scientific Reports
|May 9, 2025
Summary
This study introduces YOLO-RC, an enhanced YOLOv8 network for road traffic object detection, significantly improving accuracy for occluded and small targets. It achieves superior performance on benchmark datasets, outperforming existing methods.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Complex road environments present challenges for object detection due to dense occlusions and small targets, leading to false positives and missed detections.
- Existing object detection models often struggle with accurately localizing and identifying objects in intricate traffic scenarios.
Purpose of the Study:
- To develop an advanced road traffic object detection network, YOLO-RC, that overcomes limitations of current models in handling occlusions and small targets.
- To enhance the feature extraction, fusion, and attention mechanisms within the YOLOv8 architecture for improved detection accuracy.
Main Methods:
- The proposed YOLO-RC network enhances the YOLOv8 backbone with MBConv modules for improved feature extraction.
- A novel C3FB structure replaces the C2f module in the neck, reducing parameters and boosting accuracy.
- Weighted Bi-directional Feature Pyramid Network (BCFPN) and CBAM module are integrated for superior feature fusion and attention.
Main Results:
- YOLO-RC achieved a mAP50 of 91.1% on the DAIR-V2X-I dataset, a 4.5% improvement, with an 81.8% recall rate.
- On the UA-DETRAC dataset, YOLO-RC reached mAP50 of 94.8% and MAP50-95 of 77.9%.
- The algorithm demonstrated superior detection accuracy compared to mainstream object detection methods.
Conclusions:
- YOLO-RC effectively addresses challenges in road traffic object detection, particularly with occluded and small targets.
- The enhanced network architecture offers significant improvements in detection accuracy and generalization capabilities.
- The proposed model represents a substantial advancement in intelligent transportation systems and autonomous driving perception.
