DOUBLE-STREAM VIDEO GENERATION METHOD BASED ON DIFFERENT FEATURE SPACES OF TEXT

摘要

A double-stream video generation method based on different feature spaces of a text, comprising: a text feature extraction process, a double-stream video generation process, and an adversarial training process. In the present invention, according to an object appearance feature space and a corresponding motion feature space thereof described in the text, by separating a space feature and a temporal feature comprised in the text, modeling these features in a double-stream manner respectively, and optimizing a generated result in an adversarial training manner to enable video fragments to be closer to reality, intelligent generation of the video fragments conforming to the text content is realized. The method has wide market demands and application prospects in the aspects of material accumulation and data set automatic generation.