KFC tracker performance disappoints
As a new user of OpenCV i thought lets have a look at the various trackers that OpenCV offers. I need to choose the best tracker for an application i am developing. I have a few tens of test videos to try each tracker with. I was surprised to see poor tracking accuracy of the KFC tracker, as i have read very positive comments about it. It has the lowest CPU load of all trackers, which is great, but it very often fails to track target objects, even when they seem 'simple' to handle. I saw a video on youtube (https://www.youtube.com/watch?v=pj-Qu...) where someone has run a variety of OpenCV trackers on the same set of video fragments, and there the KFC came out as one of the most reliable. This video is a few years old. Could this be a regression in OpenCV?
I am using OpenCV 3.4.1, compiled with gcc on a mac mini running OS X 10.10.
if you have groundtruth data (bounding boxes for each frame) for your testvideos, you should check the benchmark prgram here
also, other tracker models, like MOSSE, GOTURN, CSRT have beeen added there in the meantime