README.md

---
license: mit
tags:
  - 3d
  - depth-estimation
  - multilayer-depth
  - point-cloud
  - diffusion
  - video
  - dynamic-scene
library_name: torch
pipeline_tag: image-to-3d
extra_gated_heading: "Request access to the World Tracing dynamic model"
extra_gated_description: >
  These checkpoints are released for research and product
  experimentation under the **MIT license**.  Please share a few
  details below so we can keep a light audit trail of how the
  weights are used in the wild.  Requests are reviewed manually,
  typically within **1-3 business days**.
extra_gated_button_content: "Submit access request"
extra_gated_fields:
  Full name: text
  Affiliation (university / company): text
  Country: country
  Primary intended use:
    type: select
    options:
      - Academic research
      - Personal / hobbyist project
      - Industrial research
      - Commercial product
      - Other
  Brief description of your intended use: text
  I agree to cite the World Tracing paper in any publication or release that uses these weights: checkbox
---

# World Tracing — Dynamic Model (16-frame video, r76)


## Access

The checkpoints in this repo are released under the **MIT license**,
but downloads are **gated** so we can keep a light audit trail of
how the model is used.  To download:

1. Scroll up and fill in the **"Submit access request"** form (basic
   contact info + a short note on intended use).
2. We review every request manually, usually within **1-3 business
   days**.  You will receive an email from Hugging Face once your
   request is approved.
3. After approval, log in with `huggingface-cli login` (or set
   `HF_TOKEN`) and run any of the inference examples from the
   [GitHub repo](https://github.com/haoz19/world-tracing) — the `wt`
   package picks the token up automatically and `--ckpt r75b` /
   `r69e` / `r76` triggers a normal `hf_hub_download`.

> *Note:* this is a **manual review** flow, not an auto-approve
> click-through.  We read every request individually, so please give
> a one-line description of what you plan to use the weights for.


EMA-only release weights for the **r76** dynamic-video model from
[*World Tracing: Generative Pixel-Aligned Geometry Beyond the Visible*](https://haoz19.github.io/world-tracing-page/).

* **Repo**: <https://github.com/haoz19/world-tracing>
* **Project page**: <https://haoz19.github.io/world-tracing-page/>
* **Config name**: `r76`
* **Architecture**: `MultilayerXYZModel` with temporal attention
  blocks, 2.1 B params
* **Input**: 16 frames × 336 × 336 RGBA (single shared crop across the
  clip)
* **Output**: per-frame, per-layer XYZ; 16 stacked time-steps × 6
  depth layers
* **Training data**: dynamic-object synthetic clips + curated
  real-world dynamic clips

## Files

| File | Size | Format |
|---|---|---|
| `model.pt` | 7.80 GB | bare `state_dict`, float32 |

EMA weights only — ~26 % of the original training checkpoint.

## Usage

```bash
git clone https://github.com/haoz19/world-tracing
cd world-tracing
pip install -e ".[viz]"

python examples/infer_video.py \
    --image_dir examples/test_images/dynamic/davis__camel/ \
    --ckpt      r76 \
    --config    r76 \
    --out       /tmp/wt_camel.rrd
```

Bare `--ckpt r76` triggers `huggingface_hub.hf_hub_download` against
this repo.  The clip directory must contain 16 frames (or pass
`--frame_indices "0,2,4,..."` to subsample).

## Citation

```bibtex
@misc{zhang2026worldtracing,
  title         = {World Tracing: Generative Pixel-Aligned Geometry Beyond the Visible},
  author        = {Hao Zhang and Mohamed El Banani and Jen-Hao Cheng and Paul Zhang
                   and Yi Hua and Ben Mildenhall and Christoph Lassner
                   and Narendra Ahuja and Gengshan Yang},
  year          = {2026},
  eprint        = {TODO},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV}
}
```

## License

MIT — see the [GitHub repo](https://github.com/haoz19/world-tracing/blob/main/LICENSE).