We provide data in a convenient format .hdf5 as well as .raw. The hdf5 format contains the followings:

  
  $ h5ls -r 250220_121732/prophesee0/events.hdf5
  /                        Group
  /raw_events              Group
  /raw_events/p            Dataset {2076982/Inf}
  /raw_events/t            Dataset {2076982/Inf}
  /raw_events/x            Dataset {2076982/Inf}
  /raw_events/y            Dataset {2076982/Inf}
  /trigger                 Group
  /trigger/t               Dataset {562/Inf}
  /trigger/type            Dataset {562/Inf}
  

• 'raw_events':
  • 'x': x coordinate (height).
  • 'y': y coordinate (width).
  • 't': Timestamp in microsec.
  • 'p': Polatiry, 0 is negative and 1 is positive.
• 'trigger':
  • 'type': Sync trigger type, 1 indicates ON.
  • 't': Timestamp in microsec.

RGB data is stored in mp4 file. The timing of the frames is the first trigger of every three ON triggers.

Annotation bounding boxes are provided as annotation.yaml in each sequence with CVAT v1.1 format. The annotation coordinates are in the event camera image plane, and (x, y) corresponds to (width, height) from the top-left. The annotations are at every ON trigger timing.

  
    <track id="180" label="LED-21" source="semi-auto" task_id="2335" subset="default">
    <box frame="33441" keyframe="1" outside="0" occluded="0" xtl="615.50" ytl="197.87" xbr="624.62" ybr="208.80" z_order="0">
    </box>
    <box frame="33442" keyframe="1" outside="0" occluded="0" xtl="616.82" ytl="200.57" xbr="625.61" ybr="210.60" z_order="0">
    </box>
    <box frame="33443" keyframe="1" outside="0" occluded="0" xtl="617.84" ytl="203.20" xbr="626.31" ybr="210.97" z_order="0">
    </box>
    <box frame="33444" keyframe="1" outside="0" occluded="0" xtl="618.59" ytl="203.60" xbr="626.74" ybr="211.10" z_order="0">
    </box>
    ...
  

The OptiTrack marker is attached on top of the stereo camera, hence there is a few (5-10) cm offset between the true pose of the cameras and the GT pose from OptiTrack. The camera pose is provided as 4-by-4 matrices (R, T) in .npz file format.

LED positions are given in references folder. Each json file stores the LED IDs in the 3D coordinate (same as the camera pose). The unit is in meter.

  
    "LED": {
      "21": [
          -8.967055952338944,
          0.8104301314442308,
          1.493524872
      ],
      ...
    }
  

Summary

The dataset link: Dropbox You can find the summary and details of each sequence below.

Translation

250220_123911 (Medium Sensitivity)

250220_124103 (Low Sensitivity)

Translation-Z

	1200lux	600lux
High	250220_123732	250220_123356
Medium	250220_123911	250220_123056
Low	250220_124103 250220_090916	250220_122758 250220_091233

250220_115427 (Medium Sensitivity)

250220_115642 (Low Sensitivity)

Translation-XYZ

	1200lux	600lux
Medium	250220_115427	250220_115217
Low	250220_115642	250220_115028

Rotation

250220_162949 (High Sensitivity)

250220_111743 (Low Sensitivity)

Rotation

	1200lux	600lux
High	250220_162949 250220_163133	250220_163411 250220_163539
Medium	250220_112109 250220_112247	250220_111141 250220_111300
Low	250220_111743 250220_111855	250220_111009 250220_111525

Dynamic

250220_144802 (Low Sensitivity)

Dynamic

	1200lux	600lux	10lux
High	250220_145238	250220_145624	250220_150946
Medium	250220_145032 250220_102247	250220_102630 250220_145829	250220_150742
Low	250220_144802 250220_102053	250220_150017 250220_102835	250220_150518

Static

Static

	1200lux	600lux
High	250220_113843 250220_113936 250220_114422 250220_114616 250220_115045 250220_115252 250220_120218	250220_120621 250220_121021 250220_121137 250220_121456 250220_121622 250220_122047
Medium	250220_113720 250220_114021 250220_114302 250220_114649 250220_114955 250220_115344 250220_120124	250220_120658 250220_120951 250220_121235 250220_121430 250220_121650 250220_122008
Low	250220_113604 250220_114058 250220_114208 250220_114735 250220_114915 250220_115700 250220_115953 250220_120024	250220_120752 250220_120913 250220_121301 250220_121403 250220_121732 250220_121935