Techniques to use a neural network to expand an image
AssigneeNVIDIA Corporation
InventorsGuilin Liu, Andrew Tao, Bryan Christopher Catanzaro, Ting-Chun Wang, Zhiding Yu, Shiqiu Liu, Fitsum Reda, Karan Sapra, Brandon Rowlett
Life of this patent
Estimating the term…
Abstract
Apparatuses, systems, and techniques for texture synthesis from small input textures in images using convolutional neural networks. In at least one embodiment, one or more convolutional layers are used in conjunction with one or more transposed convolution operations to generate a large textured output image from a small input textured image while preserving global features and texture, according to various novel techniques described herein.
Claims21
1-20. (canceled)
One or more processors, comprising:
- circuitry to:
- cause one or more neural networks to generate one or more first feature maps corresponding to a first image;
- cause the one or more neural networks to generate one or more second feature maps by weighting the one or more first feature maps;
- combine the one or more first feature maps with the one or more second feature maps based, at least in part, on the weighting;
- and
- generate a second image based, at least in part, on the one or more combined feature maps, wherein the second image has a higher resolution than a resolution of the first image.
A system, comprising:
- one or more processors to:
- cause one or more neural networks to generate one or more first feature maps corresponding to a first image;
- cause the one or more neural networks to generate one or more second feature maps by weighting the one or more first feature maps;
- combine the one or more first feature maps with the one or more second feature maps based, at least in part, on the weighting;
- and
- generate a second image based, at least in part, on the one or more combined feature maps, wherein the second image has a higher resolution than a resolution of the first image.
A method, comprising:
- causing one or more neural networks to generate one or more first feature maps corresponding to a first image;
- causing the one or more neural networks to generate one or more second feature maps by weighting the one or more first feature maps;
- combining the one or more first feature maps with the one or more second feature maps based, at least in part, on the weighting;
- and
- generating a second image based, at least in part, on the one or more combined feature maps, wherein the second image has a higher resolution than a resolution of the first image.
Dependencies are inferred from the claim text.