acm-header
Sign In

Communications of the ACM

ACM TechNews

Making Computers Smarter--with TikTok Dance Videos


View as: Print Mobile App Share:

TikTok dance videos usually only feature one person and show their entire body in multiple poses, making them the perfect dataset for computer vision research.

Credit: Yasamin Jafarian

The University of Minnesota's Yasamin Jafarian is using TikTok dance videos to build lifelike three-dimensional (3D) avatars of people for use in virtual reality environments.

Jafarian said such technology is expensive and mainly used by the film industry, and she "wanted to generate the same opportunity for the average person so they can just use their phone camera and be able to create a 3D avatar of themselves."

Her algorithm requires just one photo or video of a person to produce a realistic avatar, and the TikTok videos provided a sufficient training dataset.

Jafarian selected 340 10- to 15-second-long videos for training, and has used the algorithm to generate a 3D avatar of a person from the front.

From University of Minnesota News and Events
View Full Article

 

Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account