Three-dimensional modeling method and device and server

Abstract

The invention provides a three-dimensional modeling method and device and a server, and the method comprises the steps: obtaining a single-view image comprising a target object, and a sampling point in a preset three-dimensional space; extracting local features of the sampling points and symmetric points of the sampling points from the single-view image, and outputting signed distance parameters of the sampling points based on the local features, wherein the sampling points and the symmetrical points are symmetrical based on a preset reference plane in a three-dimensional space; and establishing a three-dimensional model of the target object according to the signed distance parameters of the sampling points. According to the method, the local features of the sampling points not only include the local features of the sampling points, but also include the local features of the symmetric points of the sampling points, so that the local features of the sampling points are more accurate andricher, and the effect of a three-dimensional model is improved; meanwhile, due to the fact that the local features of the symmetrical points of the sampling points are added, even if the sampling points are shielded, the accurate local features of the sampling points can be obtained through the symmetrical points, and the accuracy of the three-dimensional model is improved.

Description

technical field [0001] The present invention relates to the technical field of three-dimensional modeling, in particular to a three-dimensional modeling method, device and server. Background technique [0002] 3D reconstruction is a classic problem in graphics; among them, in the 3D reconstruction method based on the explicit domain, the 3D grid information is directly predicted from the input single-view image; while in the 3D reconstruction method based on the implicit domain, the input The single-view image and the given sampling point let the model learn the implicit value of the 3D model of the object in the image corresponding to the sampling point in the 3D space, and then obtain the implicit domain of the 3D model of the object, and finally obtain the final reconstruction through the implicit domain 3D model of the object. [0003] In related technologies, the 3D reconstruction method based on the implicit domain can be realized by OccNet (OccupancyNetworks, Occupan...

AI Technical Summary

Problems solved by technology

[0003] In related technologies, the 3D reconstruction method based on the implicit domain can be realized by OccNet (OccupancyNetworks, Occupancy Network) and DISN (Deep Implicit Surface Network, Deep Implicit Surface Network); however, affected by the limitation of viewing angle, single-view image The shape representation of an object is ambiguous. For example, different shapes may have the same shape in a single-view image, and the same shape may have different shapes in a single-view image under different viewing angles; these lead to reconstruction based on a single-view image. The effect of the 3D model is poor; in addition, there is a self-occlusion problem in the single-view image. For example, if the front of the object is displayed in the single-view image, the front of the object will block the back of the object, so the single-view image cannot be displayed at the same time. The front and back of the object; thus making it difficult to collect accurate feature points in the occluded area, which will lead to errors in the 3D model

Images