US 11,657,255 B2
Controlling a neural network through intermediate latent spaces
Sylvain Philippe Paris, Jamaica Plain, MA (US); Erik Andreas Härkönen, Helsinki (FI); and Aaron Phillip Hertzmann, San Francisco, CA (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Feb. 21, 2020, as Appl. No. 16/798,263.
Prior Publication US 2021/0264234 A1, Aug. 26, 2021
Int. Cl. G06N 3/04 (2006.01); G06N 3/08 (2023.01)
CPC G06N 3/0454 (2013.01) [G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
1. In a digital content creation digital medium environment, a method implemented by at least one computing device, the method comprising:
receiving first data for a generator network of a generative adversarial network (GAN), the generator network having multiple layers including an initial layer, a first layer and a second layer;
receiving an input selection of a first effect for a new image being generated by the generator network;
generating second data by modifying the first data based on the input selection, the modifying comprising applying decomposition vectors to activation values generated by the first layer;
providing the second data to the second layer, the second layer being a later layer in the generator network than the first layer; and
generating, using the generator network with the second data, the new image with the first effect.