Sharp-webgpu
This repository contains a staged ONNX export of Apple's SHARP monocular
view-synthesis model. It is a derived artifact of the official
apple/Sharp checkpoint.
This is not the official Apple checkpoint repository and is not endorsed by
Apple. The package contains model graph files, external tensor data, checksums,
licenses, and a manifest describing the staged artifact layout.
Intended Use
- Non-commercial scientific research and academic development.
- Evaluation of a staged ONNX derivative of the released SHARP research model.
- Integration into research clients that implement the required SHARP
preprocessing, staged inference, and postprocessing steps.
Not Intended For
- Commercial exploitation.
- Product development.
- Use in a commercial product or service.
- Any use outside the Apple Machine Learning Research Model License Agreement.
License
The source code in the upstream SHARP project is covered by LICENSE.
The model weights and derived model artifacts in this repository are covered by
LICENSE_MODEL, the Apple Machine Learning Research Model License Agreement.
That license limits use, modification, derivatives, and redistribution to
research purposes and excludes commercial exploitation, product development, or
commercial product/service use.
This ONNX artifact bundle is a derived artifact from the released SHARP model and
should be treated as a model derivative for licensing purposes.
Attribution required by LICENSE_MODEL:
Apple Machine Learning Research Model is licensed under the Apple Machine
Learning Research Model License Agreement.
Artifact modifications from the upstream apple/Sharp release:
- Exported the PyTorch checkpoint to a staged ONNX layout.
- Split execution into
stage-1 through stage-4.
Citation
If you use this model, cite the upstream SHARP work:
1@article{mescheder2025sharp,
2 title = {Sharp Monocular View Synthesis in Less Than a Second},
3 author = {Lars Mescheder and Wei Dong and Shiwei Li and Xuyang Bai and Marcel Santos and Peiyun Hu and Bruno Lecouat and Mingmin Zhen and Amael Delaunoy and Tian Fang and Yanghai Tsin and Stephan R. Richter and Vladlen Koltun},
4 journal = {arXiv preprint arXiv:2512.10685},
5 year = {2025},
6 url = {https://arxiv.org/abs/2512.10685},
7}
