README.md

---
title: Apple SHARP ZeroGPU
emoji: 🧊
colorFrom: blue
colorTo: gray
sdk: gradio
sdk_version: 6.19.0
python_version: 3.12.12
app_file: app.py
short_description: Single-image SHARP to 3DGS PLY on ZeroGPU.
models:
  - IdlecloudX/ml-sharp-weights
tags:
  - zero-gpu
  - gradio
  - 3d
  - gaussian-splatting
  - monocular-view-synthesis
preload_from_hub:
  - IdlecloudX/ml-sharp-weights sharp_2572gikvuh.pt
---

# Apple SHARP ZeroGPU

This Space wraps [apple/ml-sharp](https://github.com/apple/ml-sharp) for a public research demo on Hugging Face ZeroGPU.

Upload a single image and the Space returns a downloadable 3D Gaussian Splatting `.ply` file. The output is a 3DGS representation, not a mesh, OBJ, or GLB file.

## Usage and License Limits

Apple SHARP model weights are released under the Apple Machine Learning Research Model License Agreement. They are limited to scientific research and non-commercial use. See the linked upstream repository and the model repository license file before using the weights or outputs.

This Space does not modify or fine-tune the Apple model. It only loads the published checkpoint and exports SHARP's predicted 3DGS scene.

## Implementation Notes

- Runtime: Gradio Space on ZeroGPU.
- GPU function: `@spaces.GPU(duration=60, size="large")`.
- Model source: `IdlecloudX/ml-sharp-weights/sharp_2572gikvuh.pt`.
- SHARP source commit: `apple/ml-sharp@1eaa046834b81852261262b41b0919f5c1efdd2e`.
- Default output: `.ply` only. Video rendering is intentionally not enabled in this first version.