DRSTNet: A Robust Spatio-temporal Network with Dilated Residual Convolutions for Lane Detection

Lane detection plays a more and more significant role in ensuring the safety of autonomous driving, Lane Departure Warning, etc. Although a lot of research has been conducted with innovative methods on lane detection, pursuing the high accuracy of lane detection in challenging scenarios is still an open research question. In this work, we present a robust lane detection model via dilated residual convolutions and spatio-temporal networks (DRSTNet). The dilated residual convolutions make our model have the ability to obtain richer and denser feature information by expanding the receptive fields of the convolutions, and provide our model with necessary supplements by skip connections. In addition, the spatio-temporal networks further enhance the learning ability of our model in extracting effective features by dealing with spatial and temporal information via convolutional gated recurrent units (ConvGRUs). Furthermore, a large number of experiments verify that our model outperforms the state-of-the-art algorithms while increasing the robustness and reducing the size of the weight parameter, achieving 81.35% on DET and 73.0% on CULane.