Shengyi Qian, Linyi Jin, Chris Rockwell, *Siyi Chen, David Fouhey*
Advisor: David Fouhey
Mentor: Shengyi Qian
Independently predict 3D partial human poses as SMPL meshes in pytorch3d following Rockwell’s methods. And use PointRend to obtain 2D human masks.
Independently predict 2D plane masks as well as 3D articulation information.
Use a differential render in pytorch3d similar to the NMR in PHOSA for back-propagation, to build a similar system optimizing the position and size of the person considering 3d space relationship and interaction for each single image.
Student Group: Andrew Keisling, Brendan Nell, Kaiwen Lu, Siyi Chen
Advisors: Chaya Norton, Paul Apisa
Mentors: Christopher Zhang, Sayantan Khan