---
license: other
license_name: nvidia-source-code-license
license_link: LICENSE
tags:
- 6d-pose-tracking
- event-camera
- novel-object
---

# Event6D Weights

Pretrained checkpoints for [Event6D](https://github.com/mickeykang16/Event6D).

| File | Size | Purpose |
|------|------|---------|
| `depth_extrapolation.pth` | 1.3 MB | Event6D depth-extrapolation model (`--depth_extrapolation_ckpt`) |
| `e2vid.pth.tar` | 41 MB | E2VID image reconstruction (`--e2vid_ckpt`) |
| `foundationpose_refiner/model.pth` | 66 MB | FoundationPose refiner backbone (`--ckpt`) |
| `foundationpose_refiner/config.yml` | <1 KB | Refiner config — auto-loaded from the same directory as `model.pth` |

## Download

```bash
huggingface-cli download mickeykang/Event6D-weights --local-dir ./weights
```

After download, your directory should look like:

```
weights/
├── depth_extrapolation.pth
├── e2vid.pth.tar
└── foundationpose_refiner/
    ├── config.yml
    └── model.pth
```

## License

This release contains weights derived from
[FoundationPose (NVIDIA)](https://github.com/NVlabs/FoundationPose) and therefore
inherits the **NVIDIA Source Code License** — non-commercial / research-use only. The
full text is shipped as `LICENSE` next to this README.

`e2vid.pth.tar` is the upstream pretrained checkpoint from
[uzh-rpg/rpg_e2vid](https://github.com/uzh-rpg/rpg_e2vid) (GPL-3.0); see that repository
for the original license. `depth_extrapolation.pth` is trained by us on top of the CBMNet
architecture ([intelpro/CBMNet](https://github.com/intelpro/CBMNet)).

By using these weights you agree to the terms of the bundled `LICENSE` file.