Reproduce “EGGHead: Efficient Generalizable Gaussian Heads from a Single Image” 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
Pilseo Park, Andrew Hou, Michel Sarkis, Ning Bi, Yiying Tong, Xiaoming Liu
EGGHead reconstructs a 3D head from one portrait using a fixed Gaussian budget allocated adaptively by image-detail density. In an FFHQ source-view ablation at 12k Gaussians, it improves PSNR and LPIPS over uniform allocation but has slightly lower SSIM. On NeRSemble, 12k Gaussians outperform a 24k baseline on PSNR and LPIPS, not SSIM; these findings are limited to those comparisons.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.