Lane detection is a fundamental key in advanced driver assistance system (ADAS) given the driver information about lane marking in real-time which serves several critical purposes essential for enhancing driving safety like Lane keeping assistance and collision avoidance. However, the complexity of the road geometry and the variability of the environment conditions, creates many challenges for this task. Existing lane detection models suffer from low feature extracting capability and poor real-time detection, tackling those challenges, this paper propose a lane detection architecture that combined Swin transformer, convolution neural network and spatial convolution network. Experience was performed on Tusimple dataset and the model achieves F1 score of 95,1%, in addition the inference speed attains 32 frames per second which make the model satisfies the requirement of real-time responsiveness and robustness for lane detection task.