METHOD AND DEVICE FOR TRAINING NEURAL NETWORK MODEL, AND METHOD AND DEVICE FOR GENERATING TIME-LAPSE PHOTOGRAPHY VIDEO

摘要

A method and device for training a neural network model, and a method and device for generating a time-lapse photography video. The method for generating a time-lapse photography video comprises: acquiring specified images; generating, according to the specified images, an image set comprising a first pre-set number of frames of specified images; and according to the image set, performing content modeling and motion state modeling on the image set by means of a pre-trained neural network model to obtain a time-lapse photography video output by the neural network model, wherein the neural network model comprises a basic network for performing content modeling on the time-lapse photography video and an optimized network for modeling a motion state of the time-lapse photography video, and is obtained by acquiring a training sample and performing training according to the training sample comprising a training video and the image set corresponding thereto. A multi-stage generative adversarial network performing sustainability optimization on a time-lapse photography video, and content modeling and motion state modeling ensure reasonable prediction and realize the gradual generation of the time-lapse photography video from rough to subtle.