Reproduce “ARG-Avatar: Animation-Rendering Guidance for Efficient Feedforward Avatar Reconstruction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jianjin Xu, Soham Dasgupta, Fernando De la Torre
ARG-Avatar reconstructs animatable 3D avatars from a few images for rendering in new poses. It uses a human rig and rendered geometry to guide attention between avatar tokens and image regions. On unseen NeuMan, six-view reconstruction reaches 30.82 dB PSNR, versus 28.09 dB for LHM++. The advantage is reported for multiview input; one-view performance is weaker.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.