Object detection algorithm for autonomous driving: Design and real-time performance analysis of AttenRetina model

Abstract

With the continuous advancement of autonomous driving technology, how to efficiently and accurately detect objects (such as pedestrians, cyclists, traffic signs, etc.) has become a core challenge to improve the safety and reliability of the system. Existing object detection models still face the problem of insufficient accuracy and robustness when dealing with complex backgrounds and occlusions. To this end, this paper proposes the AttenRetina object detection model for autonomous driving, which combines the multi-scale feature fusion module (FPN) and the attention mechanism to significantly improve the detection ability of the model in various scenarios. Experimental results show that AttenRetina performs well on the KITTI and MS COCO datasets, and significantly outperforms other mainstream models in key indicators such as Precision, Recall and mAP. The mAP on the KITTI dataset reaches 0.86, which is more than 12% higher than the basic model, showing its great potential in autonomous driving object detection. The research in this paper provides an effective solution to the object detection problem in autonomous driving systems, and provides an important reference for future algorithm optimization and application.