Method and device for generating independent layer generative adversarial network

• Main Authors: JI XIANGYANG, ZOU QIRAN
• Format: Patent
• Language: Chinese; English
• Published: 29.07.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention discloses a method and device for generating an independent layer generative adversarial network, and the method comprises the steps: inputting a continuous random variable into a background generator and a foreground mask generator, obtaining a background layer, a foreground layer and a corresponding mask layer, and inputting the background layer, the foreground layer and the corresponding mask layer into a layer independence device, obtaining an optimization target of the layer independent device to the layer generator; and respectively training the layer generator, the discriminator and the layer independent device, and iteratively alternately training the layer generator, the discriminator and the layer independent device until an iteration ending condition is met, thereby obtaining a finally trained layer generative adversarial network. Therefore, the technical problems that content confusion between different layers is caused when the layer content is generated based on a foreground transl