event_camera_tools

This repository holds ROS/ROS2 tools for displaying and converting event_camera_msgs under ROS and ROS2. These messages are produced by the metavision_driver.

Supported platforms

Currently tested on Ubuntu 20.04/22.04 under ROS Noetic and ROS2 Galactic/Humble.

How to build

Set the following shell variables:

repo=event_camera_tools
url=https://github.com/ros-event-camera/${repo}.git

and follow the instructions here

Republish conversion nodelet

The republish node converts event_camera_msgs to legacy formats like dvs, prophesee, and decoded ("mono") event_camera messages. Note that this nodelet will be consuming a significant amount of CPU resources and should not be run unnecessarily while recording data. The following command will start a conversion nodelet to republish events from /event_camera/events to /event_camera/republished_events and /event_camera/republished_triggers (similar syntax under ROS2, see launch file for remapping):

roslaunch event_camera_tools republish_nodelet.launch camera:=event_camera message_type:=event_packet

ROS1/ROS2 Tools

All tools are provided in ROS and ROS2 (syntax below is for ROS):

• rosrun event_camera_tools echo [-b <bag>] <topic>: displays messages in event_camera_msgs format, optionally from a bag file. Example output:

  -------------------------------
  res:  640  height:  480 enc: evt3
  header stamp: 1664227781775114816
  time base:        0
  seqno:   129213
  ---
  6851488000 edge: 1  id:  6
  6851510000  390  223 1
  6851520000   33  326 0
  ...
  

• rosrun event_camera_tools perf <topic>: Sample output:

  msgs:   219.48/s drp:  0 del: 13.72ms drft: 0.0033s ev:   0.0823 M/s %ON:  46 tr:  1758.38 1/s %UP:  50
  msgs:   249.01/s drp:  0 del:  4.35ms drft: 0.0027s ev:   0.8497 M/s %ON:  52 tr:  1999.56 1/s %UP:  49
  

  The meaning of the fields is as follows:

  • msgs message rate per seconds
  • drp number of drops per second (skip in sequence numbers)
  • del delay: average time difference between message header stamp and arrival time. This includes the delay due to the driver aggregating messages.
  • drft accumulated (from start of "perf") drift between message header stamp and sensor-provided time.
  • ev event rate in millions/sec
  • %ON ratio of ON events to total (ON + OFF) events
  • tr rate of trigger messages
  • %UP ratio of UP trigger edges to total (UP + DOWN)

• rosrun event_camera_tools sync_test <cam_0_event_topic> <cam_1_event_topic> The output gives the average sensor time difference and how many samples where counted:

  avg sensor diff:  0.00846s, count:   360
  avg sensor diff:  0.00377s, count:   466
  ...
  

• rosrun event_camera_tools bag_to_raw -t <topic> -b <bag_name> -o <outout_raw_file> -c <camera_type>

  Converts bags with evt3 event_camera_msgs to raw file. The camera_type argument is necessary to produce a valid header for the raw file.

• rosrun event_camera_tools raw_to_bag -t <topic> -b <bag_file> -i <input_raw_file> -w <sensor_width> -h <sensor_height> -B <buffer_size>

  Converts raw file into bag with evt3 event_camera_msgs. The buffer size determines the size and number of ROS messages in the bag.

• rosrun event_camera_tools movie_maker -f <fps> -b <bag_name> -t <topic>

  produces sequence of frame images.

• rosrun event_camera_tools legacy_to_bag -b <input_bag_file> -o <output_bag_file> -t <topic1> [topic2 topic3 ...]

  Converts bags with DVS or Prophesee messages to evt3 event_camera_msgs.

ROS2 only tools

• ros2 run event_camera_tools find_trigger_events -i input_bag -t topic_with_trigger_events

  This tool is useful when aligning reconstructed image frames for datasets that use a pulse-per-second synchronization scheme. The find_trigger_events tool finds the ROS time and sensor time of the first trigger pulse in a rosbag. Output example:

  first trigger ROS    time: 1702577874035860000
  first trigger sensor time: 24809136000
  num triggers: 1712
  avg time between triggers:
  ROS time:    0.0499567s
  sensor time: 0.0499567s
  processed 202 number of messages
  

  The ROS and sensor times of the first trigger event serve as inputs for image reconstruction tools such as bag_to_frames.

• ros2 run event_camera_tools plot_events -b name_of_bag -o name_of_plot_file -t topic

  Creates file with 4 columns (time, x, y, polarity) for plotting. Also useful for clear text viewing of events in bag file.

License

This software is issued under the Apache License Version 2.0.