Abstract: A three-dimensional (3D) model generation method includes obtaining, based on a missing region of an initial 3D model of a target object, a camera pose sequence for performing photographing on the target object, where the initial 3D model is obtained by performing 3D modeling based on a first image sequence, and the first image sequence is obtained by performing photographing around the target object; obtaining a second image sequence, where the second image sequence is an image sequence obtained by performing photographing on the target object based on the camera pose sequence; and filling the missing region of the initial 3D model based on the second image sequence to generate a target 3D model.

Abstract: An object reconstruction method includes obtaining a plurality of target images and real poses corresponding to a case in which a camera captures the target images. Based on the target images and the real poses, a sparse point cloud whose size is consistent with a real size of a target object is constructed and a three-dimensional (3D) model of the target object is generated based on the sparse point cloud. A distribution size of the sparse point cloud generated based on the first poses is consistent with the real size of the target object such that the reconstructed 3D model is more realistic, and does not include a size distortion.

Abstract: A method includes obtaining, by a camera, a current image, determining a pattern of the target object in the current image, determining, based on the pattern of a target object in the current image, whether a three-dimensional model of the target object is reconstructible, and displaying first prompt information on a display in response to determining that the three-dimensional model of the target object is reconstructible. The current image is at least one of a plurality of frames of images of the target object. The first prompt information is usable to indicate that the three-dimensional model of the target object is reconstructible.