README.md

---
license: cc-by-nc-4.0
tags:
- vision-action
- embodied-ai
- game-dataset
- imitation-learning
- pretraining
- OWA
- mediaref
task_categories:
- robotics
---

# D2E-480p

[Project Page](https://worv-ai.github.io/d2e/) · [Paper (arXiv)](https://arxiv.org/abs/2510.05684) · [GitHub](https://github.com/worv-ai/D2E) · [OWA Toolkit Documentation](https://open-world-agents.github.io/open-world-agents/)

This is the dataset for [**D2E: Scaling Vision-Action Pretraining on Desktop Data for Transfer to Embodied AI**](https://worv-ai.github.io/d2e/). **268.7 hours** of synchronized video, audio, and input events from **29 PC games** across diverse genres (FPS, open-world, sandbox, and more), for training vision-action models and game agents.

**What's included:**

- **Video + Audio**: H.264 encoded at 480p 60fps with game audio. Fixed 0.5s keyframe intervals and disabled B-frame for efficient random seek without sequential decoding.
- **Input events**: Keyboard (press/release + key state), mouse (clicks, screen coordinates, raw HID deltas, button state), and active window info—all with nanosecond timestamps synchronized to video frames.
- **[OWAMcap](https://open-world-agents.github.io/open-world-agents/data/getting-started/why-owamcap/) format**: Built on [MCAP](https://mcap.dev/) (widely adopted in robotics). Indexed for fast random access, crash-safe writes, and standardized message schemas that work across different datasets without custom parsing.

**Recommended for:** Training game agents with vision-action trajectories, pretraining vision-action models for transfer to embodied AI (robotic manipulation, navigation), or world model / video generation training (use [D2E-Original](https://huggingface.co/datasets/open-world-agents/D2E-Original) for FHD/QHD).

> ⚠️ **2026/01/07**: We’ve completed fixes for issues introduced during reprocessing/re-filtering after applying privacy filtering (e.g., timestamps, MKV file paths). If you downloaded the data before this date, please re-download the updated version.

## Visualize

Explore recordings directly in your browser with synchronized keyboard/mouse overlay: **👉 [Open in Dataset Visualizer](https://huggingface.co/spaces/open-world-agents/visualize_dataset?repo_id=open-world-agents/D2E-480p)**

<img src="https://github.com/open-world-agents/owa-dataset-visualizer/blob/main/.github/assets/viewer.png?raw=true" alt="Dataset Visualizer Preview" width="600">

## Load the data

Install `mcap-owa-support` (OWAMcap reader), `owa-msgs` (message type definitions), and `huggingface_hub`:

```bash
pip install mcap-owa-support owa-msgs huggingface_hub
```

Then load and iterate through the data:

```python
from huggingface_hub import hf_hub_download
from mcap_owa.highlevel import OWAMcapReader

# Download a sample recording (mcap + video)
_kw = dict(repo_id="open-world-agents/D2E-480p", repo_type="dataset")
hf_hub_download(**_kw, filename="Apex_Legends/0805_01.mkv")
mcap_file = hf_hub_download(**_kw, filename="Apex_Legends/0805_01.mcap")

with OWAMcapReader(mcap_file) as reader:
    # Load a video frame
    for msg in reader.iter_messages(topics=["screen"]):
        screen = msg.decoded
        screen.resolve_relative_path(mcap_file)
        frame = screen.load_frame_array()  # numpy array (H, W, 3)
        break

    # Read keyboard events
    for msg in reader.iter_messages(topics=["keyboard"]):
        print(msg.decoded)  # KeyboardEvent(event_type='press', vk=87)
        break

    # Read raw mouse events
    for msg in reader.iter_messages(topics=["mouse/raw"]):
        print(msg.decoded)  # RawMouseEvent(last_x=12, last_y=-3, button_flags=0)
        break
```

**Learn more:** [OWAMcap format guide](https://open-world-agents.github.io/open-world-agents/data/technical-reference/format-guide/)

**For training:** We provide [owa-data](https://github.com/open-world-agents/open-world-agents/tree/main/projects/owa-data), a data pipeline that converts this dataset into HuggingFace Datasets ready for **PyTorch DataLoader**. It handles tokenization and sequence packing out of the box—so you can start training immediately without writing custom data loading code.

## Structure

Each game folder contains paired `.mcap` + `.mkv` files:

```
Apex_Legends/
├── 0805_01.mcap    # Timestamped events + frame references
├── 0805_01.mkv     # Video + audio (480p 60fps, H.264)
├── 0805_02.mcap
├── 0805_02.mkv
└── ...
```

The `.mcap` file stores lightweight [MediaRef](https://github.com/open-world-agents/MediaRef) pointers to video frames instead of raw pixels—frames are decoded on-demand from the `.mkv` when you call `load_frame_array()`. MCAP files contain timestamped messages on these topics:

| Topic            | Message Type             | Description                                                                                                                 |
| ---------------- | ------------------------ | --------------------------------------------------------------------------------------------------------------------------- |
| `screen`         | `desktop/ScreenCaptured` | Frame timestamp + MediaRef pointer to video                                                                                 |
| `keyboard`       | `desktop/KeyboardEvent`  | Key press/release with [virtual key code](https://learn.microsoft.com/en-us/windows/win32/inputdev/virtual-key-codes)       |
| `keyboard/state` | `desktop/KeyboardState`  | Currently pressed keys                                                                                                      |
| `mouse`          | `desktop/MouseEvent`     | Mouse clicks and screen coordinates                                                                                         |
| `mouse/raw`      | `desktop/RawMouseEvent`  | Raw HID movement ([→ why raw?](https://open-world-agents.github.io/open-world-agents/env/plugins/desktop/#raw-mouse-input)) |
| `mouse/state`    | `desktop/MouseState`     | Current position and button state                                                                                           |
| `window`         | `desktop/WindowInfo`     | Active window title, rect, and handle                                                                                       |

## Games

Genres: FPS (Apex Legends, PUBG), open-world (Cyberpunk 2077, GTA V), simulation (Euro Truck Simulator 2), sandbox (Minecraft), roguelike (Brotato, Vampire Survivors), and more. 29 games released (267h) from 31 games collected (335h) after privacy filtering.

| Game                   | Hours |
| ---------------------- | ----: |
| Apex Legends           | 25.6 |
| Euro Truck Simulator 2 | 19.6 |
| Eternal Return         | 17.1 |
| Stardew Valley         | 14.6 |
| Cyberpunk 2077         | 14.2 |
| MapleStory Worlds      | 14.1 |
| Rainbow Six            | 13.7 |
| Grand Theft Auto V     | 12.9 |
| Slime Rancher          | 10.7 |
| Dinkum                 | 10.4 |
| Medieval Dynasty       | 10.9 |
| Raft                   | 10.8 |
| Counter-Strike 2       | 9.9 |
| Satisfactory           | 9.8 |
| Grounded               | 9.7 |
| Ready Or Not           | 9.6 |
| Barony                 | 9.3 |
| Core Keeper            | 8.9 |
| Minecraft              | 8.6 |
| Monster Hunter Wilds   | 7.9 |
| Brotato                | 6.0 |
| PUBG                   | 4.9 |
| Vampire Survivors      | 2.8 |
| Battlefield 6          | 2.2 |
| Skul                   | 2.0 |
| PEAK                   | 1.8 |
| OguForest              | 0.8 |
| Super Bunny Man        | 0.7 |
| VALORANT               | 0.3 |

For FHD/QHD resolution, see [D2E-Original](https://huggingface.co/datasets/open-world-agents/D2E-Original).

## Citation

```bibtex
@article{choi2025d2e,
    title={D2E: Scaling Vision-Action Pretraining on Desktop Data for Transfer to Embodied AI},
    author={Choi, Suhwan and Jung, Jaeyoon and Seong, Haebin and Kim, Minchan and Kim, Minyeong and Cho, Yongjun and Kim, Yoonshik and Park, Yubeen and Yu, Youngjae and Lee, Yunsung},
    journal={arXiv preprint arXiv:2510.05684},
    year={2025}
}
```