Note
This is a Hugging Face dataset. For large datasets, ensure huggingface_hub>=1.1.3 to avoid rate limits. Learn more in the Hugging Face integration docs.
TartanGround → FiftyOne (Native Multimodal MCAP)#

Six trajectories from theairlabcmu/TartanGround, one per environment (AbandonedFactory, CyberPunkDowntown, GreatMarsh, Hospital, JapaneseCity, NordicHarbor), converted to native multimodal MCAP episodes. Each episode carries the front camera, its segmentation stream, per-frame lidar point clouds, ego pose, and IMU plot channels on a 10 Hz frame clock.
Installation#
pip install fiftyone
Usage#
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
dataset = fouh.load_from_hub(
"Voxel51/TartanGround",
name="TartanGround",
persistent=True,
)
fo.launch_app(dataset)
What you get#
6
.mcapepisodes of 757 to 3,727 framesStreams per episode:
/front-camera(JPEG),/front-segmentation(PNG),/lidar(point clouds),/ego-pose,/imu.plotPer-episode fields:
environment,trajectory,num_frames,duration
License & attribution#
The source dataset is released by the CMU AirLab under CC-BY-4.0; this subset is distributed under the same license. Changes from the source: trajectory subsetting, conversion to MCAP, and JPEG transcoding of the RGB frames.
Citation#
@article{patel2025tartanground,
title={TartanGround: A Large-Scale Dataset for Ground Robot Perception and Navigation},
author={Patel, Manthan and Yang, Fan and Qiu, Yuheng and Cadena, Cesar and Scherer, Sebastian and Hutter, Marco and Wang, Wenshan},
journal={arXiv preprint arXiv:2505.10696},
year={2025}
}