DEVICE AND METHOD FOR RECONSTRUCTING THREE-DIMENSIONAL HUMAN POSTURE AND SHAPE MODEL ON BASIS OF IMAGE

摘要

Provided are a device and a method for reconstructing a three-dimensional human posture and shape model on the basis of an image. A method according to an embodiment of the present invention comprises: encoding an input image into a latent area in which an image characteristic is included, by using an encoding network; converting the encoded latent area into a latent area in which geometric information is included; and estimating a three-dimensional model of the input image by using the converted latent area. By the method, three-dimensional human posture and shape information can be estimated from a latent area through unsupervised learning without a need to use other intermediate medium expression characteristics.