Method for generating face 3D image based on deep learning algorithm

The invention belongs to the technical field of face modeling, and relates to a method for generating a face 3D image based on a deep learning algorithm, and the method comprises the steps: employing a variational auto-encoder and a convolutional neural network, and constructing a face 3D image generation model: employing an encoder of the variational auto-encoder to compress input data into low-dimensional feature vectors for representation, the decoder acts the extracted feature vectors on the FLAME reference face model so as to generate a target 3D face model; constructing a training model, and training a face 3D image generation model based on the face 2D image data set; a user uploads a human face 2D image, the user image is reconstructed through the trained human face 3D image generation model, and finally a detailed human face 3D image of the user is obtained. According to the invention, a real three-dimensional image can be automatically generated through a single face 2D image, and the method is more