README.md

---
license: apache-2.0
task_categories:
  - robotics
  - reinforcement-learning
tags:
  - aerial-navigation
  - world-model
  - embodied-ai
  - uav
---

# ANWM-Dataset

Training / evaluation trajectories for **ANWM** (Aerial Navigation World Model),
released with the paper
[Aerial World Model for Long-horizon Visual Generation and Navigation in 3D Space](https://arxiv.org/abs/2512.21887).

- **Code**: [https://github.com/EmbodiedCity/ANWM.code](https://github.com/EmbodiedCity/ANWM.code)
- **Model**: [EmbodiedCity/ANWM](https://huggingface.co/EmbodiedCity/ANWM)

## Contents

Sharded tar archives of AirVLN-16 style trajectories (`airvln_16-*.tar`).
Each archive preserves the original folder layout (`{ID}_processed/...` with
images and `traj_data.pkl`).

## Quick start

```bash
git clone https://github.com/EmbodiedCity/ANWM.code.git
cd ANWM.code

huggingface-cli download EmbodiedCity/ANWM-Dataset --repo-type dataset \
  --local-dir data/airvln_16_shards
mkdir -p data/airvln_16
for f in data/airvln_16_shards/airvln_16-*.tar; do
  tar -xf "$f" -C data/airvln_16
done
```

After extract, trajectory directories are directly under `data/airvln_16/`.

Also download the released checkpoint:

```bash
huggingface-cli download EmbodiedCity/ANWM 0200000.pth.tar \
  --local-dir logs/anwm_cdit_airvln/checkpoints
```

See the code repository README for installation, inference, and evaluation.

## Related

- Code: https://github.com/EmbodiedCity/ANWM.code
- Model: https://huggingface.co/EmbodiedCity/ANWM
- Dataset: https://huggingface.co/datasets/EmbodiedCity/ANWM-Dataset
- Paper: https://arxiv.org/abs/2512.21887