A Lightweight Model for Road Sign Detection Based on Improved-YOLOv8
Bingxuan Wang, Gengzhao Wang · 2024
In recent years, with the continuous development of autonomous driving technology, target detection algorithms have also begun to be used in road environment detection of autonomous driving technology. However, this application scenario puts forward higher requirements on the real-time performance of the algorithm due to its particularity. To solve this problem, we created a lightweight model based on the improved YOLOv8 to detect common traffic signs on the road. Compared with the popular YOLOv8n, this model uses the PP-LCNet module to replace the original backbone network of the YOLO algorithm, which not only greatly reduces the amount of parameters and floating point accuracy required for the entire model, but also takes into account both inference speed and model performance. In order to prove the reliability and effectiveness of the model, this article compares it with YOLOv8n through experiments. The experimental results show that the floating point operation times and training time of the model are reduced by 36.6% and 43.3% respectively, while the performance of the original model is almost guaranteed. This model greatly improves the real-time performance of the model while ensuring model performance, and is expected to be applied in autonomous driving scenarios.