计算机科学
人气
人工智能
动画
计算机动画
运动(物理)
性格(数学)
虚拟现实
空格(标点符号)
人机交互
虚拟机
计算机图形学(图像)
操作系统
几何学
社会心理学
数学
心理学
标识
DOI:10.1109/raai52226.2021.9507778
摘要
Nowadays, Virtual YouTuber with its unique characteristics has achieved high popularity in the world. But at the same time, the technology used by the Virtual YouTuber also has a lot of room to improve. For example, today's Live2D technology can't show all the details of characters and perform all kinds of actions just like 3D models, when the cost of 3D modeling of both characters and backgrounds is much higher than 2D. If artificial intelligence technology can be applied to Virtual YouTuber, it is possible to solve these problems. Therefore, this paper explores the possibility of applying artificial intelligence technology to Virtual YouTuber from three aspects: character, motion capture and environment. This paper first proposes to use PIFuHD to transform 2D characters into 3D, then discusses how to use a single camera to achieve motion capture, and finally studies the stylization of background animation. Although there is still a certain distance to realize real low-cost Virtual YouTuber, with the progress of technology, there is still a lot of space to apply artificial intelligence to this industry.
科研通智能强力驱动
Strongly Powered by AbleSci AI