YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
TripoSR ONNX Exports for On-Device Image-to-3D
This repository contains ONNX exports of the TripoSR model for running image-to-3D inference entirely on mobile devices.
What is TripoSR?
TripoSR is a fast feed-forward image-to-3D model developed by Tripo AI and Stability AI. The original code and weights are available at:
- GitHub: https://github.com/VAST-AI-Research/TripoSR
- Hugging Face: https://huggingface.co/stabilityai/TripoSR
Files in this repository
| File | Description | Approx. size |
|---|---|---|
image_to_triplanes.onnx |
Image encoder + transformer that outputs triplane features | ~1.6 GB |
mlp.onnx |
Decoder MLP that maps triplane features to density and color | ~170 KB |
mlp_quantized.onnx |
INT8 dynamically quantized version of mlp.onnx |
~67 KB |
Purpose
These ONNX files are used by the mobai-assist Android application to convert
a single RGB image into a 3D mesh directly on the device. The conversion
pipeline (triplane sampling, marching cubes, and mesh post-processing) is
implemented in Kotlin/Android.
License
The original TripoSR code is released under the MIT License by Tripo AI & Stability AI. This repository redistributes ONNX conversions of the original weights and does not claim ownership of the model or the original license. Please refer to the original TripoSR repository for the full license and any model-specific usage terms.