Reproduce “Thingi10K: A Dataset of 10,000 3D-Printing Models” 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
Qingnan Zhou, Alec Jacobson
Thingi10K provides a resource for testing geometry-processing methods for 3D printing on Thingiverse models. The authors collected 10,000 models from staff-featured designs and measured mesh properties so researchers can select test inputs by geometric conditions. In this featured sample, 45% contain self-intersections and would be excluded by methods requiring intersection-free inputs; this measures eligibility, not algorithm failure or printing outcomes.
Key insights
- Models intended for real-world 3D printing frequently contain geometric imperfections such as self-intersections and non-manifoldness, which are often absent from traditional academic benchmarks.
- Existing datasets like MPZ14 are "unrealistically pristine," while ShapeNetCore often includes models with poor quality unsuitable for direct fabrication, highlighting the unique representational gap Thingi10K fills.
- User-generated contextual data (tags, design tools) combined with geometric properties can provide valuable insights into model characteristics, design practices, and their implications for printability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.