A video generation method for two non-adjacent images based on deep learning

Abstract

The invention discloses a video generation method of two non-adjacent images based on deep learning, which belongs to the field of confrontational learning and video generation, including performing linear interpolation processing on two non-adjacent images to obtain N frames of input images, and converting the N frames The input image is input to the first generator to obtain N frames of blurred video images between two frames of non-adjacent images; the N frames of video images are input to the trained second generator to obtain new N frames of clear video images, and Two frames of non-adjacent images are concatenated with new N frames of video images to generate a video. Among them, use the full convolutional layer to construct the first deep self-encoding convolutional network, use confrontation training to obtain the trained first generator, use the full convolutional layer and perform cross-layer connections to construct the second deep self-encoding convolutional network, Using confrontation training, a trained second generator is obtained. The video generated by the invention has good quality and long time.

Description

technical field [0001] The invention belongs to the field of confrontational learning and video generation, and more specifically relates to a method for generating video of two non-adjacent images based on deep learning. Background technique [0002] The prediction of video generation has always been a difficult problem in the field of computer vision. It is difficult for traditional non-deep learning algorithms to generate continuous high-quality video, but in fact video generation and prediction can be used in many fields, such as behavior analysis, intelligent monitoring, Video forecasting, animation production and more. [0003] In the 1980s, Yuan Lecun and others had proposed the basic theory of deep learning, but the hardware level used at that time could not meet its computing requirements, so the development of artificial intelligence was slow, but with the improvement of hardware level, deep learning With the rise of the , the method of using the features learned ...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the above defects or improvement needs of the prior art, the present invention provides a video generation method based on deep learning for two non-adjacent images, thereby solving the problem of poor quality and short time of generated or predicted video in the prior art. technical problem

Images