Real- Time Lane Detection Based on a Light-Weight Model in the Wild

Lane detection plays an important role in both advanced driver assistance system and autonomous vehicle domains. Benefited from the recent progress of deep learning methods, lane detection becomes more and more powerful. However, these methods usually require a high amount of numerical computation, and the performance of the algorithm always suffer from low speed due to the essential constraint of computing resources. In this paper, we propose a light-weight model which can simultaneously detect lanes in complex environments. Furthermore, a hierarchical feature fusion mechanism is proposed to refine detection module by producing high-level feature representations that are amenable to capture both rich object context and high-resolution details. Experiment results show that our proposed method achieves comparable performance with the state-of-the-art methods, with significantly improved computational efficiency.