---
license: apache-2.0
tags:
- depth-estimation
- monocular-depth
- diffusion
pipeline_tag: depth-estimation
---

# Pixel-Perfect-Depth — MoGe2 checkpoint (mirror)

This is an **unmodified mirror** of the `ppd_moge.pth` checkpoint from
[Pixel-Perfect Depth](https://github.com/gangweix/pixel-perfect-depth) (NeurIPS 2025),
the PPD variant that uses **MoGe2** semantics and delivers a ~20–30% improvement on
zero-shot benchmarks over the DA2 variant.

It is rehosted here **only** because the original file is distributed via Google Drive,
which is unreliable for automated downloads in the
[ComfyUI-PixelPerfectDepth](https://github.com/PozzettiAndrea/ComfyUI-PixelPerfectDepth)
integration. All credit belongs to the original authors.

## Source & attribution

- **Original repo:** https://github.com/gangweix/pixel-perfect-depth (Apache-2.0)
- **Original weights:** Google Drive — file id `1tabmcsbRVDKDfmO4KU1vOjurzN-wp0HV`
  (linked from the upstream README, "PPD / MoGe2" row)
- **Paper:** https://arxiv.org/abs/2510.07316

This mirror is unmodified and redistributed under the upstream **Apache-2.0** license
(see `LICENSE`). No endorsement by the original authors is implied.

## Usage

This checkpoint requires the MoGe2 encoder weights
([`moge2.pt`](https://huggingface.co/Ruicheng/moge-2-vitl-normal)) at load time, as in
the upstream `run.py --semantics_model MoGe2`.

```bibtex
@article{xu2025pixel,
  title={Pixel-perfect depth with semantics-prompted diffusion transformers},
  author={Xu, Gangwei and Lin, Haotong and Luo, Hongcheng and others},
  journal={arXiv preprint arXiv:2510.07316},
  year={2025}
}
```