GHARP divides avatar creation into two stages.
Structured reconstruction priors make Gaussian head animation faster and steadier
GHARP separates offline identity construction from lightweight runtime animation, using a fixed-size Gaussian representation that does not grow with the number of reference images.
Big Tech
Ali Benlalah · Sepehr Johari · Patricia Vitoria · Armin Kappeler · Artem Sevastopolsky · Alexander Jung · +11 more
Apple
Research Digest··3 min read
Benlalah et al.
Why this paper
From Apple
In one line
GHARP animates 3D head avatars in real time from a few input images and an expression signal by predicting expression-dependent residuals on a canonical Gaussian representation built offline.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ✓Compute or model size stated (hardware A100)
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§