The video2cfd source code in this repository is released under the
MIT License (see LICENSE). It depends on, but does not redistribute, the
third-party components below. Each is obtained from its own source and is
governed by its own license.
| Component | Used in | License | Source |
|---|---|---|---|
| SAM 3 (Segment Anything Model 3) | Step 1 (segmentation) | SAM License (Meta) — source-available, not OSI | https://github.com/facebookresearch/sam3 |
| Meshroom / AliceVision | upstream reconstruction (equirectangular→pinhole) | MPLv2 / others | https://alicevision.org |
| COLMAP | upstream reconstruction (SfM) | BSD | https://colmap.github.io |
Nerfstudio (nerfacto) |
upstream reconstruction (NeRF + point-cloud export) | Apache-2.0 | https://nerf.studio |
| OpenFOAM | downstream CFD (outside this repo) | GPLv3 | https://openfoam.org |
Important: SAM 3 is governed by Meta's SAM License, which is not a standard open-source license. It is a runtime dependency of Step 1 only and is not included here. Review and comply with its terms separately. Steps 2–5 do not require SAM 3.
numpy, pandas, scipy, networkx, open3d, opencv-python, pillow, trimesh, PyYAML, torch (and, optionally, manifold3d) — each under its own permissive license.
The repository bundles a small set of STL template meshes used for template-based
placement and mannequin insertion (chair_*.stl, table_*.stl,
person_sitting_1.stl, teacher.stl). These models were obtained from publicly
available online CAD repositories that provide them as free-to-download assets,
and are redistributed here for convenience. If you are a rights holder of any
bundled model and have concerns about its inclusion, please contact the authors.