Naoki Nozawa, Hubert P. H. Shum, Edmond S. L. Ho, Shigeo Morishima
3D Car Shape Reconstruction from a Single Sketch Image
MIG2019: Best Second Poster
Efficient car shape design is a challenging problem in both the au-tomotive industry and the computer animation/games industry. Inthis paper, we present a system to reconstruct the 3D car shapefrom a single 2D sketch image 윈도우7 무비 메이커. To learn the correlation between 2Dsketches and 3D cars, we propose a Variational Autoencoder deepneural network that takes a 2D sketch and generates a set of multi-view depth & mask images, which are more effective representationcomparing to 3D mesh, and can be combined to form the 3D carshape Buddy Rush. To ensure the volume and diversity of the training data,we propose a feature-preserving car mesh augmentation pipelinefor data augmentation. Since deep learning has limited capacityto reconstruct fine-detail features, we propose a lazy learning ap-proach that constructs a small subspace based on a few relevant carsamples in the database Download Japanese performing arts. Due to the small size of such a subspace,fine details can be represented effectively with a small number ofparameters. With a low-cost optimization process, a high-qualitycar with detailed features is created catia v5r20 다운로드. Experimental results showthat the system performs consistently to create highly realisticcars of substantially different shape and topology, with a very lowcomputational cost 베리즈 다운로드.