Communications of the ACM
CS Student Uses TikTok Dance Videos and AI to Generate 3D Avatars
Each video in the dataset depicts dynamic movements of a single person while lacking 3D ground truth geometry.
Credit: Yasamin Jafarian
University of Minnesota computer science Ph.D. student Yasamin Jafarian has been using TikTok dance videos as food for a computer algorithm that uses the frame-by-frame data to construct lifelike 3D avatars of real people.
Jafarian needed a large dataset of videos to "train" the algorithm. TikTok dance videos — which often feature one person showing the full length of their body in multiple poses — provided the perfect solution. She used 340 videos in her dataset, each one 10-to-15 seconds long. At a video rate of 30 frames per second, she had amassed more than 100,000 images of people dancing.
So far, Jafarian has been able to successfully use her algorithm to generate a 3D avatar of a person from the front view. Her published paper on the subject, co-authored by Assistant Professor Hyun Soo Park, won a Best Paper Honorable Mention Award at the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition.
From University of Minnesota
View Full Article
No entries found