RepATT: Re-parameterization and Attention-guided Hybrid Anchor Classification for Real-time Lane Detection

Lane detection is a critical technology in autonomous driving, which requires both high detection accuracy and real-time efficiency, especially in complex scenarios like severe occlusions, and dazzling light conditions. To achieve the above requirements, based on re-parameterization and attention mechanism we propose a real-time lane detection network with a classification-based manner named RepATT. A new backbone Extended RepVGG (ExtRep) which subtly decouples the training time and inference time is proposed as the feature extractor. Specifically, the ExtRep has two additional branches of varying complexities and scales compared to the original RepVGG. This enhancement improves the performance without any additional inference cost by enhancing the representational capacity of single convolution with re-parameterization methods. In addition, by integrating the Convolutional Block Attention Module(CBAM), our model improves the performance in the no-visual evidence scenarios. CBAM generates the attention maps via channel and spatial dimensions enabling the model to concentrate more on the lanes. Through comprehensive experiments conducted on CULane and TuSimple datasets, the results demonstrate our proposed method achieves the fastest inference speed, which surpasses the baseline model by 24% to 363 FPS with lower computing cost, while the F1-measure of RepATT also outperforms the previous methods on both datasets.

I. Introduction

With the development of deep learning, significant success has been achieved in autonomous driving. As a part of the environment perception module, lane detection has progressively emerged as a pivotal research focus, which is widely applied in lane-keeping assistance and adaptive cruise control. To ensure the reliability of lane detection, both high accuracy and realtime efficiency are required. However, in real-world scenarios, many challenges are encountered: severe occlusions, dazzling light, and other no-visual evidence conditions [1], which severely influence detection accuracy. Besides, the computing devices in vehicles are unable to support the complex lane detection algorithm, resulting in low inference speed.