Inputs:
- ROS bag file with several vehicles' trajectories (on
/<veh_name>/pose
topics) - video from a stationary camera
- homography matrix btwn camera plane and planar vehicle pose frame
Outputs:
- video with trajectories overlayed on the stationary camera's stream
Clone the repo, catkin_make
, then:
roslaunch video_overlay video_overlay_node.launch experiment:=4_hexes_parallel_swaps_1_pass
Set/update the args in video_overlay_node.launch
to point to the correct yaml files (homography, experiment)
If you don't have a homography matrix already, you should:
- find a recognizable moment in your video & take a screenshot
- find the corresponding vicon coordinates during that moment (from the bag)
- Click on the corresponding vehicle positions in the screenshot, using:
rosrun video_overlay video_overlay_node <screenshot_filename>
- Create a new
.yaml
file inconfig/homography_pts
with the vicon and img coords - Switch which
.yaml
file is loaded in the launch file - Your homography matrix will be computed for you at the start of the script
The outcpp.avi
file will be giant. You can compress it with ffmpeg (increase -crf to make file smaller):
ffmpeg -i outcpp.avi -crf 25 outcpp_small.avi
Contact: mfe@mit.edu