Augmented reality object rendering based on camera quality
AssigneeSnap Inc.
InventorsSergey Demyanov, Yunqing Hu, Istvan Marton, Daniil Ostashev, Aleksei Podkin
Life of this patent
Estimating the term…
Abstract
Systems and embodiments herein describe an augmented reality (AR) object rendering system. The AR object rendering system receives an image, generates a set of noise parameters and a set of blur parameters for the image using a neural network trained on a paired dataset of images, identifies an AR object associated with the image, modifies the AR object using the set of noise parameters and the set of blur parameters, displays the modified augmented reality object within the image.
Claims20
A method comprising:
- receiving, from one or more image sensors of a computer device, an image;
- generating a set of noise parameters and a set of blur parameters for the image using a neural network trained to:
- receive a first dataset of clear images and a second dataset of noisy images;
- for each clear image in the first dataset of clear images and for each noisy image in the second dataset of noisy images:
- estimate random gaussian noise of the noisy image;
- generate a modified clear image, the modified clear image generated by applying the estimated random gaussian noise to the clear image;
- and
- compare the modified clear image with the noisy image using a noise discriminator network;
- identifying an augmented reality object associated with the image;
- modifying the augmented reality object using the set of noise parameters and the set of blur parameters;
- and
- causing display, on a graphical user interface of the computer device, of the modified augmented reality object within the image.
A computing system comprising:
- a processor;
- and
- a memory storing instructions that, when executed by the processor, configure the system to:
- receive, from one or more image sensors of a computer device, an image;
- generate a set of noise parameters and a set of blur parameters for the image using a neural network trained to:
- receive a first dataset of clear images and a second dataset of noisy images;
- for each clear image in the first dataset of clear images and for each noisy image in the second dataset of noisy images:
- estimate random gaussian noise of the noisy image;
- generate a modified clear image, the modified clear image generated by applying the estimated random gaussian noise to the clear image;
- and
- compare the modified clear image with the noisy image using a noise discriminator network;
- identify an augmented reality object associated with the image;
- modify the augmented reality object using the set of noise parameters and the set of blur parameters;
- and
- cause display, on a graphical user interface of the computer device, of the modified augmented reality object within the image.
A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
- receive, from one or more image sensors of a computer device, an image;
- generate a set of noise parameters and a set of blur parameters for the image using a neural network trained to:
- receive a first dataset of clear images and a second dataset of noisy images;
- for each clear image in the first dataset of clear images and for each noisy image in the second dataset of noisy images:
- estimate random gaussian noise of the noisy image;
- generate a modified clear image, the modified clear image generated by applying the estimated random gaussian noise to the clear image;
- and
- compare the modified clear image with the noisy image using a noise discriminator network;
- identify an augmented reality object associated with the image;
- modify the augmented reality object using the set of noise parameters and the set of blur parameters;
- and
- cause display, on a graphical user interface of the computer device, of the modified augmented reality object within the image.
Dependencies are inferred from the claim text.