	US 11,748,932 B2
	Controllable image generation
Marek Adam Kowalski, Cambridge (GB); Stephan Joachim Garbin, London (GB); Matthew Alastair Johnson, Cambridge (GB); Tadas Baltrusaitis, Cambridge (GB); Martin De la Gorce, Cambridge (GB); Virginia Estellers Casas, Cambridge (GB); and Sebastian Karol Dziadzio, Cambridge (GB)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jun. 29, 2020, as Appl. No. 16/915,863.
Claims priority of provisional application 63/015,846, filed on Apr. 27, 2020.
Prior Publication US 2021/0335029 A1, Oct. 28, 2021
Int. Cl. G06T 15/00 (2011.01); G06T 9/00 (2006.01)

CPC G06T 15/005 (2013.01) [G06T 9/002 (2013.01)]

20 Claims

1. A method comprising:

storing a real image of an object in memory, the object being a type of object;

computing, using a first encoder, a factorized embedding of the real image;

receiving a value of a parameter of a synthetic image rendering apparatus, the parameter controlling an attribute of a synthetic image of an object of the type rendered by the rendering apparatus;

computing a predicted embedding factor for the received value using a second encoder;

modifying the factorized embedding including swapping at least part of the factorized embedding with the computed embedding factor; and

computing, using a decoder with the modified embedding as input, an output image of the object which is substantially the same as the real image except for the attribute controlled by the parameter.