Accurate and efficient road scene segmentation techniques are increasingly needed in many applications such as Advanced Driver Assistance System(ADAS)and autonomous driving.Conventional techniques heavily rely on handcrafted features,which limit machine learning algorithms to achieving a better performance.Recent advances of deep learning reveal that deep neural network is a promising method to satisfy the demand.Despite tremendous progress in deep learning,it still remains many open problems to be studied.This paper compares several commonly-used deep learning methods applied to road scene segmentation,which combines with the complex road traffic situations as in China.To evaluate the performance of these approaches,comparative experiments with several scenarios were performed.The accuracy and efficiency of segmentation can be obtained from the experiment and some drawbacks are shown.For these shortcomings,several preliminary suggestions for improvement are put forward to provide a potential reference for the future research of ADAS and autonomous driving.