Dataset¶
Together with ViPE, we release a large-scale dataset containing around 1M high-quality videos with accurate camera poses and dense depth maps.
| Dataset Name | # Videos | # Frames | Hugging Face Link | License | Prefix |
|---|---|---|---|---|---|
| Dynpose-100K++ | 99,501 | 15.8M | Link | CC-BY-NC 4.0 | dpsp |
| Wild-SDG-1M | 966,448 | 78.2M | Link | CC-BY-NC 4.0 | wsdg |
| Web360 | 2,114 | 212K | Link | CC-BY 4.0 | w360 |
Download datasets with:
# Replace YOUR_PREFIX with a prefix from the table.
# More specific prefixes, such as wsdg-003e2c86, download a specific shard.
uv run python scripts/download_dataset.py --prefix YOUR_PREFIX --output_base YOUR_OUTPUT_DIR --rgb --depth
Note
The depth component is very large and can take a long time to download. For the RGB component of Dynpose-100K++, ViPE retrieves frames from YouTube. Install yt_dlp and ffmpeg-python to use that path. See the original Dynpose-100K dataset for alternative retrieval approaches.
Visualize downloaded dataset artifacts with:
uv run vipe visualize YOUR_OUTPUT_DIR