Human body action migration image generation method based on nerve radiation field

The invention relates to the field of computer graphics and deep learning, in particular to a human body action migration image generation method based on a neural radiation field, which comprises the following steps: preprocessing a source human body action image and a target action semantic mask, and inputting a trained NeRF decoupling generator G, obtaining the probability S, the density sigma and the color C of the semantic category to which each pixel point output by the G belongs; generating a synthetic human body image, a synthetic semantic mask and a reverse semantic mask based on the semantic category probability S, the density sigma and the color C; adaptively adjusting the parameter of the G according to the synthetic semantic mask and the reverse semantic mask; and obtaining a synthesized human body image of the finally output human body action migration image again. According to the method, the problem of mutual adhesion of shapes and appearances in the action migration process of an existing met