FNon R-CNN: A multi-scale ground object detection and recognition network
Zhuo Yan, Wenyuan Zhu, Xinyu Zhong, Deyuan Zhang, Chuanyun Wang, Linlin Wang, Xiaocong Zhang, Feng Han · High-Confidence Computing · 2025
With the rapid development of smart cities and intelligent transportation, traffic issues such as congestion and accidents have become increasingly critical. This paper addresses the challenge of multi-scale object detection in complex scenes by proposing an improved Faster R-CNN model named FNon R-CNN. The model enhances global context modeling through integrating Non-Local Blocks into the backbone network, achieves multi-level feature fusion via Feature Pyramid Network, and optimizes training with a dynamic loss function. Experimental results on the SODA-D dataset demonstrate significant improvements in detection accuracy, particularly for small objects.