syguan96/DynaBOA
[T-PAMI 2022] Out-of-Domain Human Mesh Reconstruction via Dynamic Bilevel Online Adaptation
This project helps motion capture professionals, animators, or anyone needing to analyze human movement by reconstructing a detailed 3D human mesh from standard 2D videos, even those captured in uncontrolled, "in-the-wild" environments. It takes video footage (from a webcam or pre-recorded file) as input and outputs a precise 3D model of the human subject's pose and shape. This is ideal for researchers studying human biomechanics or animators generating realistic digital avatars.
228 stars. No commits in the last 6 months.
Use this if you need to accurately reconstruct 3D human body shapes and poses from videos shot with varying camera angles, lighting, and subject movements, especially when existing models struggle with your diverse footage.
Not ideal if you require real-time 3D reconstruction without any prior setup or if you only need basic 2D pose estimation rather than a full 3D mesh.
Stars
228
Forks
19
Language
Python
License
—
Category
Last pushed
Aug 30, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/syguan96/DynaBOA"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
tamasino52/U-CondDGCN
Unofficial pytorch implementation of U-CondDGCN from "WenBo Hu, Changgong Zhang, Fangneng Zhan,...
UPC-ViRVIG/SparsePoser
Official Code for ACM Transactions on Graphics (TOG) paper "SparsePoser: Real-time Full-body...
Silverster98/HUMANISE
Official implementation of the NeurIPS22 paper "HUMANISE: Language-conditioned Human Motion...