Code for Self-Supervised Representation Learning for CAD
Depends on AutoMate, CMake Parasolod, and OpenCascade. To setup Parasolid dependency, set the $PARASOLID_BASE
environmental variable as described here.
We evaluate against Fusion 360 Segmentation, MFCAD, and FabWave datasets for segmentation and classification tasks.
Run the download.py
script to obtain these datasets (they should not be unzipped). FabWave requires contacting the authors to ask for a download link.
The experimental data for all of our experiments is available here. Plots from the paper can be reproduced with the generate_figures.py
script.
If you use our work, please cite us as:
TBD (In Submission)