Generative neural network distillation
AssigneeSnap Inc.
InventorsSergey Tulyakov, Sergei Korolev, Aleksei Stoliar, Maksim Gusarov, Sergei Kotcur, Christopher Yale Crutchfield, Andrew Wan
Life of this patent
Estimating the term…
Abstract
A compact generative neural network can be distilled from a teacher generative neural network using a training network. The compact network can be trained on the input data and output data of the teacher network. The training network train the student network using a discrimination layer and one or more types of losses, such as perception loss and adversarial loss.
Claims19
A system comprising:
- one or more processors of a machine;
- and
- one or more memories storing instructions that, when executed by the one or more processors, cause the machine to perform operations comprising:
- identifying an object in an image;
- selecting a neural network trained to apply an effect to images that comprise the object;
- and
- generating a result image by using the neural network to apply the effect to the image, wherein the neural network is trained on input data and output data generated by a teacher generative neural network, and wherein the neural network is trained based on default losses for the effect and an additional loss specific to the object.
A method comprising:
- identifying, by one or more processors, an object in an image;
- selecting a neural network trained to apply an effect to images that comprise the object;
- and
- generating a result image by using the neural network to apply the effect to the image, wherein the neural network is trained on input data and output data generated by a teacher generative neural network, and wherein the neural network is trained based on default losses for the effect and an additional loss specific to the object.
A non-transitory machine-readable storage medium embodying instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
- identifying, by one or more processors, an object in an image;
- selecting a neural network trained to apply an effect to images that comprise the object;
- and
- generating a result image by using the neural network to apply the effect to the image, wherein the neural network is trained on input data and output data generated by a teacher generative neural network, and wherein the neural network is trained based on default losses for the effect and an additional loss specific to the object.
Dependencies are inferred from the claim text.