ComfyUI-SECADNET
SECAD-Net: Sketch-Extrude CAD Reconstruction - Originally from ComfyUI-CADabra
Nodes (5)
Load SECAD-Net, then hope the Google Drive download actually lands
Turn any 3D mesh into a colored voxel block and watch it render in your browser
The node whose display name says '(Required!)' — and it's not lying
Voxels in, rough CAD mesh out (and a latent you must keep)
Fine-tuned latent in, final CAD mesh out
Work in Progress! This node is not finished.
ComfyUI-SECADNET
Sketch-extrude CAD reconstruction from voxel representations.
Originally from ComfyUI-CADabra
Paper
SECAD-Net: Self-Supervised CAD Reconstruction by Learning Sketch-Extrude Operations
- GitHub: https://github.com/BunnySoCrazy/SECAD-Net
Installation
Via ComfyUI Manager
Search for "SECADNET" in ComfyUI Manager
Manual Installation
cd ComfyUI/custom_nodes
git clone https://github.com/PozzettiAndrea/ComfyUI-SECADNET
pip install -r ComfyUI-SECADNET/requirements.txt
Nodes
- LoadSECADNetModel - Load pretrained encoder/decoder/generator
- MeshToVoxel - Convert mesh to 64x64x64 voxel grid
- SECADNetInference - Initial reconstruction from voxels
- SECADNetFinetune - Per-shape optimization (200-500 epochs)
- SECADNetLatentToMesh - Generate mesh from latent codes
Requirements
- torch>=2.0.0
- numpy>=1.24.0
- trimesh>=3.20.0
- PyMCubes>=0.1.4
- gdown>=4.7.0
Community
Questions or feature requests? Open a Discussion on GitHub.
Join the Comfy3D Discord for help, updates, and chat about 3D workflows in ComfyUI.
Credits
- Original CADabra: PozzettiAndrea/ComfyUI-CADabra
- SECAD-Net paper authors
License
GPL-3.0