Reproduce “Sprint3D: Accelerating End-to-End 3D Asset Generation through Pipeline Optimization” 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
Sho Kuno, Ruihan Yu, Yonghao Yu, Guixu Lin, Lian Fu, Muyao Niu, Yu-Ju Tsai, Zheng-Hui Huang and 6 more
Sprint3D speeds the released Pixal3D image-to-3D pipeline by replacing CPU mesh processing with GPU processing and shortening its four flow stages. It lays mesh surfaces into 2D UV atlases for texture baking. On one H200, median end-to-end time falls from 133.0 seconds to 4.05 seconds, 33×, under official export settings. Normal-map fidelity improves, but mean silhouette IoU is slightly lower.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.