CODEWITHSUNDEEP
algorithmcomputer-vision
Vadim Smolyakov
Vadim Smolyakov

Reputation: 1197

high performance object trackers in computer vision

I'm interested in tracking multiple objects in a video using a (potentially moving) single camera in an environment with partial occlusions and variable distance from objects to the camera. I was wondering what are some of the good tracking algorithms that can solve this problem? I experimented with mean and cam-shift as well as background subtraction but wasn't able to get good results.

Upvotes: 1

Views: 160

Answers (1)

serycjon
serycjon

Reputation: 540

I would recommend reading through results of VOT challenge 15. And then continue researching selected trackers. http://www.votchallenge.net/vot2015/

Correlation filter based trackers (like SRDCF - https://www.cvl.isy.liu.se/research/objrec/visualtracking/regvistrack/ ) achieve very high speed (due to using Fast Fourier Transform), state-of the-art results and are not that hard to implement, so that is where I would look first. (SRDCF original implementation is single-target, but it is easy to extend it by running multiple instances)

Upvotes: 1

Related Questions