ViCTer: A semi-supervised video character tracker
The paper, ViCTer: A semi-supervised video character tracker, was made available on-line on March 11, 2023 on Science Direct in the Machine Learning With Applications Journal. MLWA is a peer reviewed, open access journal focused on research related to machine learning.
It was written by Zilinghan Li, Xiwei Wang, (students in CAII) and Zhenning Zhang along with Dr. Volodymyr Kindratenko, Director, Center for AI Innovation at NCSA.
Some highlights include:
- Semi-supervised learning boosts accuracy in video-analysis problems.
- Triplet loss helps face recognition networks better distinguish unseen faces.
- Combining face recognizer and human tracker solves video character tracking problems.
To read more of the paper, go here: https://doi.org/10.1016/j.mlwa.2023.100460