With the rethinking of modernist urban planning and delicacy management of urban construction, higher requirements are put forward for the vitality evaluation of urban streets. It is difficult for present evaluation index of urban vitality to relate the layout of streets to people's feelings. So the paper combines people's intuitive feelings of the street in the evaluation of the vitality of the city, Further More, The method based on deep learning will be more objective to the evaluation of indicators. Firstly, the paper use the FCN network to semantically segment street scenes and analysis the component of the segmented image layout, then the composition ratio of each street element are extracted to get the evaluation of the street scale, building density and so on. Secondly, a large number of facial expressions are used in the pre-training of CNN, and then the paper use opencv to detect the face in the original street pictures and use trained CNN to analysis the emotion of the face to get people's intuitive feelings about the street. Lastly, the comprehensive evaluation of the vitality of th e city will be gave by combining many other present evaluation factors. However, this method mentioned require the images keep high visual quality and it has limitation of the number of classification of the emotion of the people's face.

CICTP