The UZH-FPV Drone Racing Dataset:

High-speed, Aggressive 6DoF Trajectories for State Estimation and Drone Racing

UZH-FPV Drone Racing Dataset

We introduce the UZH-FPV Drone Racing dataset, which is the most aggressive visual-inertial odometry dataset to date. Large accelerations, rotations, and apparent motion in vision sensors make aggressive trajectories difficult for state estimation. However, many compelling applications, such as autonomous drone racing, require high speed state estimation, but existing datasets do not address this. These sequences were recorded with a first-person-view (FPV) drone racing quadrotor fitted with sensors and flown aggressively by an expert pilot. The trajectories include fast laps around a racetrack with drone racing gates, as well as free-form trajectories around obstacles, both indoor and out. We present the camera images and IMU data from a Qualcomm Snapdragon Flight board, ground truth from a Leica Nova MS60 laser tracker, as well as event data from an mDAVIS 346 event camera, and high-resolution RGB images from the pilot's FPV camera. With this dataset, our goal is to help advance the state of the art in high speed state estimation.


IROS 2019 FPV Drone Racing VIO Competition

We are organizing the first FPV Drone Racing VIO competition using this dataset. The competition is held jointly with IROS 2019 Workshop "Challenges in Vision-based Drone Navigation". The goal is to estimate the quadrotor motion as accurately as possible, utilizing any desired sensor combinations. The winner will be be awarded 1,000 USD and will also be invited to present his approach at the workshop.

The submission deadline is Sunday, September 1st, 2019. Details of the competition can be found here.


Citing

When using the data in an academic context, please cite the following paper.
Information field illustration

J. Delmerico, T. Cieslewski, H. Rebecq, M. Faessler, D. Scaramuzza

Are We Ready for Autonomous Drone Racing? The UZH-FPV Drone Racing Dataset

IEEE International Conference on Robotics and Automation (ICRA), 2019.

PDF YouTube Project Webpage and Datasets


Visualization of Dataset Sequences



Dataset Format

We provide all datasets in two formats: text files and binary files (rosbag). While their content is identical, some of them are better suited for particular applications. The binary rosbag files are intended for users familiar with the Robot Operating System (ROS) and for applications that are intended to be executed on a real system.



Binary Files (rosbag)

The rosbag files contain images and IMU measurements using the standard sensor_msgs/Image and sensor_msgs/Imu message types, respectively. The events are provided as dvs_msgs/EventArray message types, and the ground truth is provided as geometry_msgs/PoseStamped messages. The Events/IMU/GT bag files also contain the image frames from the mDAVIS as sensor_msgs/Image messages.


Datasets

Indoor forward facing

# Duration
(s)
Length
(m)
vmax
(m/s)
DAVIS Snapdragon Ground
Truth
Public*
Rosbag ZIP Rosbag ZIP
3 54.63 287.12 9.5 853.7MB 413.9MB 1.6GB 842.8MB yes
Easy
5 50 156.47 4.87 1.3GB 655.3MB 2.6GB 1.3GB yes
Medium
6 32.93 223.27 12.52 728.1MB 365.1MB 1.2GB 605.3MB yes
Medium
7 73.2 333.59 12.78 1.1GB 559.5MB 2GB 964.1MB yes
Hard
8 132.53 259.16 5.26 1.4GB 726.9MB 2.8GB 1.4GB no
9 34.04 157.07 11.42 710.5MB 365.6MB 1.3GB 627MB yes
Easy
10 33.43 149.36 9.49 643.7MB 327.9MB 1.3GB 678.7MB yes
Easy
11 24.02 85.68 10.32 973.2MB 467.5MB 1.4GB 731.3MB no
12 31.98 124.07 15.28 565.1MB 268.4MB 1GB 567.8MB no
* Some ground truth data is withheld for the Drone Racing VIO competition at IROS 2019 workshop "Challenges in Vision-based Drone Navigation" in Macau.


Indoor 45 degree downward facing

# Duration
(s)
Length
(m)
vmax
(m/s)
DAVIS Snapdragon Ground
Truth
Public*
Rosbag ZIP Rosbag ZIP
1 73.99 150.71 4.36 1.3GB 635.8MB 1.6GB 897.3MB no
2 55.77 218.9 6.97 1.3GB 602.5MB 1.2GB 643.5MB yes
Easy
3 57.82 119.82 3.53 870.6MB 437.8MB 1.3GB 683.1MB no
4 47.36 168.06 6.55 1.1GB 496.3MB 1.1GB 607.8MB yes
Easy
9 40 215.58 11.23 808.1MB 370.8MB 1.2GB 664MB yes
Medium
11 22.96 125.21 11.74 949.6MB 440.1MB 893.3MB 480.8MB no
12 51.25 124.56 4.33 563.7MB 268.8MB 1.3GB 648.6MB yes
Easy
13 42.49 166.62 7.92 527.9MB 254.7MB 1.1GB 534.9MB yes
Medium
14 43.66 220.4 9.54 625.6MB 301MB 1.2GB 543.8MB yes
Hard
16 15.49 58.72 7.69 336MB 166.1MB 805.3MB 448MB no
* Some ground truth data is withheld for the Drone Racing VIO competition at IROS 2019 workshop "Challenges in Vision-based Drone Navigation" in Macau.


Outdoor forward facing

# Duration
(s)
Length
(m)
vmax
(m/s)
DAVIS Snapdragon Ground
Truth
Public*
Rosbag ZIP Rosbag ZIP
1 49.63 258.23 8.55 687.2MB 358.6MB 1.5GB 664.8MB yes
Easy
2 36.9 220.88 10.13 763.4MB 410.3MB 249.8MB 710.4MB no
3 92.84 735.51 14.04 1.1GB 637.2MB 2.2GB 1.1GB yes
Medium
5 22.21 189.63 20.73 657.1MB 342.7MB 1.5GB 729.6MB yes
Hard
6 34.83 338.2 19.42 786.5MB 427MB 1.3GB 677.6MB no
9 43.15 314.41 10.68 1.1GB 577.4MB 1.5GB 742.1MB no
10 59.6 455.63 12.58 1.2GB 656MB 1.8GB 955.6MB no
* Some ground truth data is withheld for the Drone Racing VIO competition at IROS 2019 workshop "Challenges in Vision-based Drone Navigation" in Macau.


Outdoor 45 degree downward facing

# Duration
(s)
Length
(m)
vmax
(m/s)
DAVIS Snapdragon Ground
Truth
Public*
Rosbag ZIP Rosbag ZIP
1 24.49 165.53 15.62 1.3GB 650.4MB 1.4GB 723.3MB yes
Medium
2 26.19 143.13 10.68 1.1GB 588.1MB 1.2GB 633.5MB no
* Some ground truth data is withheld for the Drone Racing VIO competition at IROS 2019 workshop "Challenges in Vision-based Drone Navigation" in Macau.

Calibration

We provide the calibration parameters for the camera intrinsics and camera-IMU extrinsics in YAML format, as well as the raw calibration sequences used to produce those with the Kalibr toolbox.




For sequences DAVIS Snapdragon
Calib. sequences Calibration Calib. sequences Calibration
Camera IMU Camera IMU
Indoor forward 134.1MB 138.6MB 2.2MB 1.2GB 1GB 4.1MB
Indoor 45 down 206.8MB 112.8MB 2.2MB 1.3GB 792.3MB 2.2MB
Outdoor forward 213.1MB 227.6MB 4.4MB 919.1MB 892.9MB 5.3MB
Outdoor 45 down 214.7MB 130.1MB 2.3MB 1.4GB 578.4MB 3.5MB



License

This datasets are released under the Creative Commons license (CC BY-NC-SA 3.0), which is free for non-commercial use (including research).

Acknowledgements

This work was supported by the National Centre of Competence in Research Robotics (NCCR) through the Swiss National Science Foundation, the SNSF-ERC Starting Grant, and the DARPA FLA Program.

This work would not have been possible without the assistance of Stefan Gächter, Zoltan Török, and Thomas Mörwald of Leica Geosystems and their support in gathering our data. Additional thanks go to Innovation Park Zürich, and the Fässler family for providing experimental space, and iniVation AG and Prof. Tobi Delbruck for their support and guidance with the mDAVIS sensors.