🤖 AI Summary
To address low detection and recognition accuracy of traffic signs under complex conditions—including small objects, occlusion, illumination variations, and class imbalance—this paper proposes a lightweight and efficient YOLOv8-based framework. The method innovatively integrates Coordinate Attention (CA), BiFPN for multi-scale feature fusion, ODConv for dynamic convolution, and LSKA for large-kernel selective attention, jointly optimized with EIoU/WIoU loss functions and Focal Loss to enhance classification discrimination. Evaluated on GTSRB, TT100K, and GTSDB benchmarks, the approach achieves consistent mAP improvements of 3.2–5.7% over baseline models, significantly boosting robustness in adverse scenarios while maintaining real-time inference capability on edge devices. This work delivers a high-accuracy, low-latency traffic sign perception solution tailored for autonomous driving and ADAS applications.
📝 Abstract
This paper Traffic sign recognition plays a crucial role in the development of autonomous vehicles and advanced driver-assistance systems (ADAS). Despite significant advances in deep learning and object detection, accurately detecting and classifying traffic signs remains challenging due to their small sizes, variable environmental conditions, occlusion, and class imbalance. This thesis presents an enhanced YOLOv8-based detection system that integrates advanced data augmentation techniques, novel architectural enhancements including Coordinate Attention (CA), Bidirectional Feature Pyramid Network (BiFPN), and dynamic modules such as ODConv and LSKA, along with refined loss functions (EIoU and WIoU combined with Focal Loss). Extensive experiments conducted on datasets including GTSRB, TT100K, and GTSDB demonstrate marked improvements in detection accuracy, robustness under adverse conditions, and real-time inference on edge devices. The findings contribute actionable insights for deploying reliable traffic sign recognition systems in real-world autonomous driving scenarios.