Jump to Content
  1. Queen's
  2. Smith Eng.
  3. ECE

Estimation, Search, and Planning (ESP) Research Group

Oxford Multimotion Dataset (OMD)

Welcome to the Oxford Multimotion Dataset (OMD). The data are hosted on Google Drive (no sign-in required) and are accessible from either the web folder or the individual file links below.

Return to Top

Full documentation of the dataset is available in the associated RA-L paper. Please cite the dataset as:

Authors
  1. K. M. Judd
  2. J. D. Gammell
Title
The Oxford Multimotion Dataset: Multiple SE(3) motions with ground truth
Publication
Journal
IEEE Robotics and Automation Letters (RA-L)
Volume
4
Number
2
Pages
800–807
Date
Notes
Presented at ICRA 2019
Data
Data
Videos
Video
PDFs
PDF
Digital Object Identifier (DOI)
doi: 10.1109/LRA.2019.2892656
arXiv
Google Scholar
Google Scholar
Return to Top

Useful scripts and tools for manipulating the data and evaluating results are provided here.

Return to Top

There are two calibrations for the dataset. All of the primary segments and most of the secondary segments with imu data use the same calibration. The secondary segments without imu data use a separate no_imu calibration.

Each calibration consists of three files:

  1. Camera intrinsics as provided by the camera manufacturer (manufacturer.yaml)
  2. Camera intrinsics & extrinsics as estimated by Kalibr using the calibration_extrinsic data segment (kalibr.yaml)
  3. Transform from the camera frame to the Vicon frame as estimated from the calibration_vicon data segment (vicon.yaml)
Return to Top

Here we list the known issues with and updates to the dataset.

  • The Vicon transforms are described incorrectly in the paper. They are stored in the CSV as passive transforms from object to world. The accompanying GitHub code is written for passive transforms from world to object and so converts the Vicon transforms on ingestion.
  • Some data segments exhibit a small amount of jitter in sections of the Vicon ground-truth data. This can be mitigated through simple interpolation and smoothing.
  • Though the stereo camera, RGB-D camera, and IMU were recorded on the same machine, they are not hardware synchronized. The Vicon was recorded on a separate system with an unknown temporal offset and clock drift.
  • Updates to dataset files:
    • 2019-06-14: Improved calibration parameters for vicon.yaml and kalibr.yaml are now included in the dataset and linked in the segments below. Previous calibrations are still available if desired.
Return to Top
Return to Top
Primary Data Segments
Segment (Duration)Preview (YouTube)DataROS DataCalibrations
swinging_4_static
6m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
swinging_4_translational
3m 45s
Watch on YouTube
stereo.tgz
rgbd.tgz
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
swinging_4_unconstrained
6m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
occlusion_2_static
8m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
occlusion_2_translational
4m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
occlusion_2_unconstrained
6m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_3_static
3m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_3_translational
2m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_3_unconstrained
6m 30s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_6_static
4m 45s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_6_translational
3m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_6_unconstrained
3m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_6_robot
(2m)
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
Return to Top
Secondary Data Segments
Segment (Duration)Preview (YouTube)DataROS DataCalibrations
calibration_extrinsic
2m 45s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
calibration_vicon
3m 15s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
calibration_extrinsic_no_imu
5m 45s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
calibration_vicon_no_imu
2m 50s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
pinwheel_1_static
1m 5s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
pinwheel_1_unconstrained
1m 15s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
fixed_occlusion_1_static
3m 25s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
fixed_occlusion_1_translational
1m 50s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
fixed_occlusion_1_unconstrained
4m 50s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_1_unconstrained
3m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
imu.csv
vicon.csv
stereo.bag
rgbd.bag
imu.bag
vicon.bag
manufacturer.yaml
kalibr.yaml
vicon.yaml
cars_2_static_no_imu
2m 20s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
cars_6_static_no_imu
10m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
cars_6_unconstrained_no_imu
5m 00s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
cars_6_robot_no_imu
6m 50s
Watch on YouTube
stereo.tgz
rgbd.tgz (raw_depth.tgz)
vicon.csv
stereo.bag
rgbd.bag
vicon.bag
manufacturer_no_imu.yaml
kalibr_no_imu.yaml
vicon_no_imu.yaml
Return to Top
  1. Authors
    1. K. M. Judd
    2. J. D. Gammell
    Title
    The Oxford Multimotion Dataset: Multiple SE(3) motions with ground truth
    Publication
    Journal
    IEEE Robotics and Automation Letters (RA-L)
    Volume
    4
    Number
    2
    Pages
    800–807
    Date
    Notes
    Presented at ICRA 2019
    Data
    Data
    Videos
    Video
    PDFs
    PDF
    Digital Object Identifier (DOI)
    doi: 10.1109/LRA.2019.2892656
    arXiv
    Google Scholar
    Google Scholar