Is there a way to optimize MultiTracker_update method?

asked 2019-04-01 11:19:51 -0600

gepi gravatar image


I am currently working on a project where a task is to update several trackers on a frame basis. To do that I am using MultiTrakcer_update method provided by cv2 api (Python stack). Object tracker is CSRT tracker. I was wondering whether there is a way to optimize it so that code runs faster. Currently, the script works the following way:

  1. I initialize several trackers using .add method. The tracker is CSRT. Unfortunately, others failed to track correctly.
  2. Trackers are updated on a frame-basis.

I was thinking of possible two solutions: 1. Initialize trackers and then use several processes to update trackers. 2. Apply some pre-processing to the frame so that trackers have less work to do.

What is your opinion?

Kind regards and have a nice day, gepi

edit retag flag offensive close merge delete


I always think CSRT is a little too slow (and it has no intrinsic boundary checking so if the tracking box is near the boundary of field of view it would crash). I like to use KCF (faster) and MOSSE ( really fast! )

gino0717 gravatar imagegino0717 ( 2019-04-01 21:01:51 -0600 )edit