Reproduce “FrameSeg3D: Learned Local Frames for Segmenting Small Parts on High-Resolution 3D Scans” 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
Xinwen Zhang, Tongyan Wang, Huawei Lin
FrameSeg3D segments small parts in detailed 3D scans by first estimating each part’s center and facing direction from a coarse scan, then rendering close-up views around it at scanner resolution. For nails, part-facing views score Dice 0.970, versus 0.965 for equally centered mat-normal views. Tests cover nails and teeth; on the separate tooth challenge set, ToothInstanceNet has higher per-tooth IoU.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.