Method and device for generating new character image from original character image

Abstract

The invention relates to the field of artificial intelligence, and particularly relates to a method for generating a new character image from an original character image which comprises the followingsteps: providing a first original character image; obtaining attitude key points of the first original character image; segmenting the first original character image into a foreground image and a background image; inputting the foreground image and the attitude key point data into a foreground feature extraction model to extract an appearance feature vector and an attitude feature vector, and inputting the background image into a background feature extraction model to extract a background feature vector; and inputting the appearance character vector, the attitude character vector and the background character vector into a synthesis model to synthesize a reconstructed image. The method also includes inputting first and second original character images to the trained neural network model tosynthesize a new character image, with the new character image having a character appearance in the first original character image and a background and a character pose in the second original character image. The invention also relates to a device for generating the new character image from the original character image.

Description

technical field [0001] The invention relates to a method for generating a new person image from an original person image. The invention also relates to a device for generating a new character image from an original character image. Background technique [0002] In recent years, with the development of artificial intelligence such as deep learning and neural network, generative models represented by confrontational generative network (Generative Adavsiral Network, GAN) and variational autoencoder (Variational Autoencoder, VAE) have made great progress. , and is widely used in the generation of data such as images and speech. [0003] In the field of image generation, the method of image generation of people is also becoming a research hotspot. However, the current character generation network / algorithm usually cuts out the image containing the person in the form of a rectangular block (bounding box or patch), and then inputs it into the network for training, and often only ...

AI Technical Summary

Problems solved by technology

However, it is difficult for existing generative networks to achieve real and natural "appearance transfer"

Images