US 12,394,149 B2
Method, electronic device, and computer program product for virtual reality modeling utilizing encoding and projection of multiple distinct views
Zhisong Liu, Shenzhen (CN); Zijia Wang, Weifang (CN); Zhen Jia, Shanghai (CN); Sanping Li, Beijing (CN); and Tianxiang Chen, Shanghai (CN)
Assigned to Dell Products L.P., Round Rock, TX (US)
Filed by Dell Products L.P., Round Rock, TX (US)
Filed on Mar. 27, 2023, as Appl. No. 18/126,619.
Claims priority of application No. 202310183674.9 (CN), filed on Feb. 28, 2023.
Prior Publication US 2024/0331276 A1, Oct. 3, 2024
Int. Cl. G06T 15/04 (2011.01); G06F 3/01 (2006.01); G06T 7/40 (2017.01); G06T 17/00 (2006.01); G06V 10/44 (2022.01)
CPC G06T 17/00 (2013.01) [G06F 3/011 (2013.01); G06T 7/40 (2013.01); G06T 15/04 (2013.01); G06V 10/44 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method for virtual reality modeling, comprising:
obtaining a first image of a real object at a first viewing angle;
obtaining, based on the first image, an initial three-dimensional model corresponding to the real object;
determining a second image of the real object at a second viewing angle different from the first viewing angle by using the initial three-dimensional model, the second image having texture characteristics of the real object; and
generating a target three-dimensional model used for virtual reality and corresponding to the real object by using the second image and the first image;
wherein determining the second image of the real object at the second viewing angle comprises encoding a plurality of different views, generated by the initial three-dimensional model at respective different viewing angles, in a view encoder, and processing the encoded views in a two-dimensional encoder coupled to the view encoder to generate respective corresponding two-dimensional views, the second image corresponding to a particular one of the two-dimensional views, the view encoder being further configured to encode camera postures of the viewing angles in association with mean and variance values of a corresponding posture distribution, the two-dimensional encoder comprising a visibility encoder and first and second convolutional neural networks, the visibility encoder and the first convolutional neural network each receiving as respective inputs an encoded view from the view encoder, the second convolutional neural network receiving as inputs the first image and an output of the first convolution neural network, the second image being generated as a function of outputs of the visibility encoder and the second convolutional neural network; and
wherein generating the target three-dimensional model comprises applying the two-dimensional views to a two-dimensional to three-dimensional projector to project the two-dimensional views into three dimensions based on one or more texture maps in forming the target three-dimensional model.