# burn_reconstruction 🕊️🔥🌌
[](https://github.com/Mosure/burn_reconstruction/actions?query=workflow%3Atest)
[](https://raw.githubusercontent.com/mosure/burn_reconstruction/main/LICENSE)
[](https://crates.io/crates/burn_reconstruction)
bevy / burn feed-forward gaussian splatting. view the [wasm example](https://mosure.github.io/burn_reconstruction/?image=re10k/0.png&image=re10k/1.png&image=re10k/2.png)

## features
- [x] multi-view images -> 3d gaussians (YoNoSplat path)
- [x] CLI export to GLB (`KHR_gaussian_splatting`)
- [x] bevy UI (native + wasm)
- [x] model auto-bootstrap + cache (parts-first burnpack)
- [ ] [bevy_synth](https://github.com/mosure/bevy_synth) integration
- [ ] [bevy_zeroverse](https://github.com/mosure/bevy_zeroverse) fine-tuning/training
## setup
### install
```bash
# cli
cargo install burn_reconstruction
# bevy app
cargo install bevy_reconstruction
```
### usage
```bash
burn_reconstruction \
--images view0.png view1.png view2.png \
--output /tmp/scene.glb
```
```bash
bevy_reconstruction -- \
--image assets/images/re10k/0.png \
--image assets/images/re10k/1.png \
--image assets/images/re10k/2.png
```
> note, input images are optional for bevy_reconstruction
## license
licensed under either of:
- Apache License, Version 2.0
- MIT license
at your option.
> note: model weights have their own license
## references
- [bevy_gaussian_splatting](https://github.com/mosure/bevy_gaussian_splatting)
- [burn_dino](https://github.com/mosure/burn_dino)
- [YoNoSplat](https://github.com/cvg/YoNoSplat)