Convert rosbag files to video files for visualization

[sbrodeur]

We can use or adapt available source codes for that: https://github.com/OSUrobotics/bag2video https://github.com/mlaiacker/rosbag2video

[sbrodeur]

Video preview is now available with commit 8c39e61

Check the videos below: visualization.zip

[sbrodeur]

There are more sensors/information that we could export to videos and display:

• Motor velocities (implemented by not added to the mosaic).
• Position in 2D space (from odometry topic).
• Battery voltage, current and charge information.
• Collision sensors (range and switch sensors).
• Audio channels as analog signals (although sound is currently part of the video).

Support is to be added in file: https://github.com/sbrodeur/ros-icreate-bbb/blob/master/src/action/scripts/record/data/visualize_hdf5.py

[sbrodeur]

Currently, the implementation is very slow with Matplotlib generating animations. This should be optimized. See: http://bastibe.de/2013-05-30-speeding-up-matplotlib.html

It seems that the encoding process is responsible for the slow processing, even with simple codec (e.g rawvideo, mjpeg) compared to H.264. Using libx264 is still the fastest (33 FPS) and leads to smaller size videos. The only way to gain speed is to parallelize computation across CPU cores, but encoding multiple videos on different threads / processes. This would be easy since we have a lot of sensors to process, and this is done sequentially right now.

[sbrodeur]

See commit f6b1679 about how to use multiprocessing and downsampling to have considerable speedup.

[sbrodeur]

See commit 6fccca0 for the last visualizations that needed to be implemented.