Abstract: Face landmark detection is one of the typical problems in computer vision, which has important influence on face 3D reconstruction, expression recognition, head pose estimation, face tracking and so on. At present, the model based on deep neural network is most popular in this field. However, as the existing key point detection deep neural network structure design getting more complex, it requires much more the computing and storage resources . In this .paper, We propose a new streamlined landmark detection network structure to replace the existing network structure. Compared with other network structures, a simple network only consists of one feature extraction module and an upsampling module made up of several deconvolution layers. In addition, we add global constraints on all key points of the face in the network structure to reduce the generation of predicted outliers. Experiments show that the detection performance of this network structure with global constraints on the 300-W data has better performance than the current typical deep neural network detection model.