Picture generation method and device, storage medium, and electronic device

An apparatus for picture generation, comprising a memory for storing computer instructions and a processor in communication with the memory, wherein, when the processor executes the computer instructions, the processor is configured to cause the apparatus to:

• obtain a plurality of sample pictures;
• perform following training steps for training an adversarial neural network model by using the plurality of sample pictures, the adversarial neural network model comprising:
• a generation network model matching the picture generation model, and a determining network model, configured to determine a generation result of the generation network model:
• train the determining network model using the generation network model until a converged determining network model is obtained;
• and
• train the generation network model by using the converged determining network model until a converged generation network model is obtained;
• in response to an output result of the converged generation network model not converging to a predetermined threshold, update the generation network model with the converged generation network model and repeating the training steps;
• in response to the output result of the converged generation network model converging to the predetermined threshold, use the converged generation network model as a picture generation model;
• obtain a source portrait picture displaying a target object;
• crop the source portrait picture to obtain a face region picture corresponding to a face of the target object excluding a hair portion;
• input the face region picture to the picture generation model to obtain an output result of the picture generation model;
• and
• generate a target portrait picture by using the output result of the picture generation model, the target portrait picture displaying a target hairstyle matching the face of the target object.

A non-transitory storage medium for storing computer readable instructions, the computer readable instructions, when executed by a processor, causing the processor to:

• obtain a plurality of sample pictures;
• perform following training steps for training an adversarial neural network model by using the plurality of sample pictures, the adversarial neural network model comprising:
• a generation network model matching the picture generation model, and a determining network model, configured to determine a generation result of the generation network model:
• train the determining network model using the generation network model until a converged determining network model is obtained;
• and
• train the generation network model by using the converged determining network model until a converged generation network model is obtained;
• in response to an output result of the converged generation network model not converging to a predetermined threshold, update the generation network model with the converged generation network model and repeating the training steps;
• in response to the output result of the converged generation network model converging to the predetermined threshold, use the converged generation network model as a picture generation model:
• obtain a source portrait picture displaying a target object;
• crop the source portrait picture to obtain a face region picture corresponding to a face of the target object excluding a hair portion;
• input the face region picture to the picture generation model to obtain an output result of the picture generation model;
• and
• generate a target portrait picture by using the output result of the picture generation model, the target portrait picture displaying a target hairstyle matching the face of the target object.