I guess in deployment time obs is not given to model as 4fps

[MilkClouds]

In process_bags.py, train data is processed as frame rate, 4.
In deployment script, observation callback is being called and main loop is called 4fps. But, in my guess, since observation callback is independent to main loop's rospy.Rate(4), observations are not collected in 4fps.

Let my know if I read in wrong way.