lidavidsh/ml-sharp
1
1---2title: Sharp Monocular View Synthesis3emoji: 📷4colorFrom: indigo5colorTo: pink6sdk: gradio7sdk_version: 5.49.18app_file: app.py9python_version: 3.1210pinned: false11license: cc-by-nc-4.012---13 14# Sharp Monocular View Synthesis in Less Than a Second15 16[](https://apple.github.io/ml-sharp/)17[](https://arxiv.org/abs/2512.10685)18 19This software project accompanies the research paper: _Sharp Monocular View Synthesis in Less Than a Second_20by _Lars Mescheder, Wei Dong, Shiwei Li, Xuyang Bai, Marcel Santos, Peiyun Hu, Bruno Lecouat, Mingmin Zhen, Amaël Delaunoy,21Tian Fang, Yanghai Tsin, Stephan Richter and Vladlen Koltun_.22 2324 25We present SHARP, an approach to photorealistic view synthesis from a single image. Given a single photograph, SHARP regresses the parameters of a 3D Gaussian representation of the depicted scene. This is done in less than a second on a standard GPU via a single feedforward pass through a neural network. The 3D Gaussian representation produced by SHARP can then be rendered in real time, yielding high-resolution photorealistic images for nearby views. The representation is metric, with absolute scale, supporting metric camera movements. Experimental results demonstrate that SHARP delivers robust zero-shot generalization across datasets. It sets a new state of the art on multiple datasets, reducing LPIPS by 25–34% and DISTS by 21–43% versus the best prior model, while lowering the synthesis time by three orders of magnitude.26 27## Getting started28 29We recommend to first create a python environment:30 31```32conda create -n sharp python=3.1333```34 35Afterwards, you can install the project using36 37```38pip install -r requirements.txt39```40 41To test the installation, run42 43```44sharp --help45```46 47## Using the CLI48 49To run prediction:50 51```52sharp predict -i /path/to/input/images -o /path/to/output/gaussians53```54 55The model checkpoint will be downloaded automatically on first run and cached locally at `~/.cache/torch/hub/checkpoints/`.56 57Alternatively, you can download the model directly:58 59```60wget https://ml-site.cdn-apple.com/models/sharp/sharp_2572gikvuh.pt61```62 63To use a manually downloaded checkpoint, specify it with the `-c` flag:64 65```66sharp predict -i /path/to/input/images -o /path/to/output/gaussians -c sharp_2572gikvuh.pt67```68 69The results will be 3D gaussian splats (3DGS) in the output folder. The 3DGS `.ply` files are compatible to various public 3DGS renderers. We follow the OpenCV coordinate convention (x right, y down, z forward). The 3DGS scene center is roughly at (0, 0, +z). When dealing with 3rdparty renderers, please scale and rotate to re-center the scene accordingly.70 71### Rendering trajectories (CUDA GPU only)72 73Additionally you can render videos with a camera trajectory. While the gaussians prediction works for all CPU, CUDA, and MPS, rendering videos via the `--render` option currently requires a CUDA GPU. The gsplat renderer takes a while to initialize at the first launch.74 75```76sharp predict -i /path/to/input/images -o /path/to/output/gaussians --render77 78# Or from the intermediate gaussians:79sharp render -i /path/to/output/gaussians -o /path/to/output/renderings80```81 82## Evaluation83 84Please refer to the paper for both quantitative and qualitative evaluations.85Additionally, please check out this [qualitative examples page](https://apple.github.io/ml-sharp/) containing several video comparisons against related work.86 87## Citation88 89If you find our work useful, please cite the following paper:90 91```bibtex92@inproceedings{Sharp2025:arxiv,93 title = {Sharp Monocular View Synthesis in Less Than a Second},94 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 Ama\"{e}l Delaunoyand Tian Fang and Yanghai Tsin and Stephan R. Richter and Vladlen Koltun},95 journal = {arXiv preprint arXiv:2512.10685},96 year = {2025},97 url = {https://arxiv.org/abs/2512.10685},98}99```100 101## Acknowledgements102 103Our codebase is built using multiple opensource contributions, please see [ACKNOWLEDGEMENTS](ACKNOWLEDGEMENTS) for more details.104 105## Web Interface (One-Click Setup)106 107For a simple web interface where you can upload images and download 3D Gaussians:108 109**macOS:** Double-click `start.command` in Finder.110 111This will automatically:112- Create the conda environment if it doesn't exist113- Install all dependencies if needed114- Start the web server at http://localhost:8000115 116117118 119 120## License121 122Please check out the repository [LICENSE](LICENSE) before using the provided code and123[LICENSE_MODEL](LICENSE_MODEL) for the released models.124