Third-party notices
September 9, 2026 ยท View on GitHub
Original EG-HumanNeRF contributions are licensed under CC BY-NC 4.0. The components below retain their upstream terms and copyright notices. Adaptations include module organization, tensor interfaces and integration with the rendering pipeline.
| Component | Source and attribution | License notes |
|---|---|---|
| VolRecon | VolRecon, copyright 2022 Fangjinhua Wang; reconstruction and camera-processing components | MIT |
| SMPL/SMPL-X code | Acquired separately from ICON's SMPL-X distribution, revision ba2ee5681284c3d627305b1c919f4414009f753b; originally developed by MPG | Not bundled. The installer preserves the upstream license and verifies file hashes. Model assets require separate access under the model terms |
| ZJU loader and keypoint image encoder | source/dataset/zju.py, source/img_encoders/kpt_cnn.py; KeypointNeRF, copyright Meta Platforms, Inc. and affiliates | CC BY-NC 4.0 |
| NVIDIA/Facebook neural image operators | source/dnnlib/, source/torch_utils/, StyleGAN-derived layers in eghumannerf/model/upsampling.py and source/legacy/upscaling_cnn.py; StyleGAN2-ADA and StyleNeRF | NVIDIA Source Code License for StyleGAN-derived code. StyleNeRF's Meta contributions use CC BY-NC 4.0, as specified in its license statement. Per-file notices apply |
| FFC | source/ffc.py, adapted from LaMa | Apache 2.0; residual-block and refinement interfaces modified |
| PyTorch3D, TorchSparse, tiny-cuda-nn, spconv | Installed dependencies; exact source revisions/distributions in pyproject.toml and uv.lock | Dependencies are installed separately under their upstream licenses |
| nvdiffrast | Optional rasterizer installed from NVlabs/nvdiffrast, pinned in pyproject.toml and uv.lock | Installed separately under its upstream license |
| pyimgui, GLFW bindings and PyOpenGL | Optional GUI dependencies: pyimgui, pyGLFW, PyOpenGL | Installed separately; each dependency's bundled license applies |
| THuman and ZJU data | THuman official access/terms, ZJU official access | Obtain data directly from the dataset owners under their terms |
| SMPL/SMPL-X models, UVs and associated topology | Body-model parameters and associated resources | Obtain separately under the applicable model licenses |
| Paper media | docs/assets/method.png from the authors' preprint figure; published article credits Wang, Kanamori and Endo and carries a CC BY 4.0 notice with normal third-party-material exceptions | Figure attribution: Wang, Kanamori and Endo; third-party credits, where present, retain their own terms |