GHARP proposes real-time 3D head animation from a few imagesMachine translation
Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.
GHARP builds a subject’s appearance offline, then uses a lightweight network to animate it from an expression signal at runtime. A body alignment network addresses missing information about body pose and clothing position. The paper reports state-of-the-art quality on Ava-256, with up to 13× faster execution on an A100 GPU and 8× fewer Gaussians.
This source permits summary display only.
Read at the original source