动漫
计算机科学
人工神经网络
人工智能
深度学习
作者
Polina Belova,Ksenia Urkaeva,Anna A. Gamova
出处
期刊:IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering
日期:2021-01-26
标识
DOI:10.1109/elconrus51938.2021.9396557
摘要
With the development of generative adversarial networks, algorithms that implement the generation of images of various topics have become widespread. This article explores the use of GAN for analysis and generation of complex animation objects and its training on a sample of Japanese animation. The visual component was chosen as the parameter of the opening ideality, which reduced the generation task to the reproduction of frames of the desired video, according to which the artist or a separate neural network would be able to continue the animated video sequence. An original dataset was assembled, consisting of 60 popular openings, and a generative-discriminative neural network created on the basis of existing analogues was trained.
科研通智能强力驱动
Strongly Powered by AbleSci AI