Reproduce “A Real-World CAD Segmentation Benchmark: Comparing Rule-Based and Learning-Based Approaches through Hole Detection” 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
Kai Schütte, Edgar Heinert, Matthias Rottmann
The authors compare two rule-based detectors with UV-Net and AAGNet for identifying functionally labelled fastener holes in engineering CAD, and introduce a manually annotated benchmark of 2,202 models. Learning-based methods perform strongly in-domain, but on a filtered 200-part MFCAD++ test, GWS achieves per-hole F1 of 0.845 versus AAGNet’s 0.026. This single transfer test does not establish universal robustness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.