Deep learning is being applied in many fields, and its application in transport engineering is becoming increasingly significant. Transport engineering needs deep learning to process traffic big data for better prediction and recognition, and ultimately better designing, planning, management, and more comfortable driving environment. This review collects many papers concerning deep learning utilized in transport engineering and the topics are categorized as three parts which are flow prediction, speed prediction and safety prediction respectively. By analyzing and comparing deep learning with other algorithms, it shows that deep learning employed in transport engineering has a good performance in prediction and recognition. However, since deep learning is immature, the deep learning used in transport engineering has some constraints and still needs further study, such as choosing the number of layers, activation function, and elements as well as framework optimization.