## 💾 Checkpoints

To keep this repository lightweight, all model checkpoints are hosted on [Hugging Face](https://huggingface.co/lingxiang2023/IBRSteG/tree/main/model_zoo). 

Please download the required weights and place them in your local `model_zoo/` directory.

| File | Purpose | Download |
| --- | --- | --- |
| `gps_plus_final.pth` | Frozen GPS-Gaussian+ backbone checkpoint | [Download](https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/gps_plus_final.pth) |
| `ibrsteg_test_weight.pth` | Inference-only IBRSteG/GAS checkpoint for testing | [Download](https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/ibrsteg_test_weight.pth) |

**Quick Download via CLI:**

```bash
mkdir -p model_zoo
wget [https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/gps_plus_final.pth](https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/gps_plus_final.pth) -O model_zoo/gps_plus_final.pth
wget [https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/ibrsteg_test_weight.pth](https://huggingface.co/lingxiang2023/IBRSteG/resolve/main/model_zoo/ibrsteg_test_weight.pth) -O model_zoo/ibrsteg_test_weight.pth
```

Note: The original full checkpoint (\~489 MB, containing optimizer/scheduler states) has been trimmed to the inference-only ibrsteg_test_weight.pth (\~163 MB, containing explicitly formulated camera parameter steganographic keys and minimal metadata) for faster evaluation.

## Links & Citation

If you find our work useful in your research, please consider citing our paper and starring the repository:

* **Paper (arXiv):** [[2606.30024] IBRSteG: Learning a Generalizable Steganography Framework for 3D Gaussian Splatting](https://arxiv.org/abs/2606.30024)
* **Code (GitHub):** [LingXiang2023/IBRSteG](https://github.com/LingXiang2023/IBRSteG)

```bibtex
@article{kong2026ibrsteg,
  title={IBRSteG: Learning a Generalizable Steganography Framework for 3D Gaussian Splatting},
  author={Kong, Fanye and others},
  journal={arXiv preprint arXiv:2606.30024},
  year={2026}
}

---
license: mit
---