Human body three-dimensional modeling system based on hand-drawn sketch

Abstract

The invention discloses a human body three-dimensional modeling system based on a hand-drawn sketch, and the system comprises an input sketch module which is used for obtaining a sketch image inputtedby a user; a deformation-based human body model generation module which is used for extracting morphological characteristics from the sketch image through a neural network and deforming the human body standard template model based on the target morphological characteristics to obtain a deformed human body three-dimensional model; and a parameterized model optimization module which is used for predicting human body posture parameters and human body shape parameters according to the deformed human body three-dimensional model and the human body standard template model so as to optimize the deformed human body three-dimensional model. According to the system, the high-quality human body three-dimensional model of the corresponding posture can be obtained through simple lines edited by a useror drawn by hand, high-precision and complex sketch input is not needed, and the system can be widely applied to various human body modeling software.

Description

technical field [0001] The invention relates to the fields of computer graphics and computer-aided design, in particular to a three-dimensional human body modeling system based on hand-drawn sketches. Background technique [0002] Sketch-based modeling is a problem of constructing corresponding 3D models from planar sketches. Before the rise of deep learning, traditional methods generally used the edge features of sketch lines to predict the shape of 3D models, for example, literature {Lipson, H. and Shpitalni, M., "Optimization-based reconstruction of a 3D object from a single freehand line drawing,” in Computer-Aided Design, 1996. Zeleznik, P.C., Herndon, K.P., and Hughes, J.F., “Sketch: An interface for sketching 3D scenes,” in Proceedings of SIGGRAPH, 1996.}. [0003] Traditional methods are only suitable for regular objects, require precise line drawings, and cannot represent the details of three-dimensional surfaces. After the rise of deep learning, researchers used ...

AI Technical Summary

Problems solved by technology

For example, the literature {Johanna, D., Mathieu, A., Phillip, I., Efros, A., and Adrien, B., “3D sketching using multi-view deep volumetric prediction,” in Proceedings of the ACM on Computer Graphics and Interactive Techniques, 2018.} Using a 3D convolution network to generate a voxel model, but because the sketch itself can provide less geometric details, and because the use of 3D convolution is greatly limited by the storage and technical capabilities of the computer, the resolution of the generated 3D model low rate

[0004] In addition, literature {Lun, Z., Gadelha, M., Kalogerakis, E., Maji, S., and Wang, R., "3Dshape reconstruction from sketches via multi-view convolutional networks," inInternationalConference on 3D Vision, 2017. }Through the deep neural network to generate the depth map and normal direction of each perspective from the sketches of multiple perspectives, and then generate a 3D model by fusing the point cloud of multiple perspectives, but there is a lack of geometric texture information in the sketch, and the accuracy of the depth map synthesized by the network is average. The final model will lack surface detail

[0005] It can be seen that the current sketch modeling method based on deep learning lacks three-dimensional surface details, lacks realism, and cannot meet the needs of many practical applications.

Images