Hi,
I am about to embark on a bit of a journey into Computer Vision. My Goal is to track a single (but changeable) Object in the CameraPreview of an Android Phone. I have done some research online about what different methods there are to achieve this, I found a lot of videos showing that it works, but sadly i didn't find any examples published on github. From what I saw and read i would implement a MOSSE tracker because it seems not to be too computationally expensive (which would mean i could run it on a phone). Additionally i was thinking i would like to combine this with a detection network which i would run sparsely on some of the Frames - depending on how much cost is asociated with that. This I would do to correct any accumulated tracking error of the bounding box which i would like to be as accurate as possible. That said this is only a plan not having done any work with CV algorithms before. I would appreciate any tips on performance of the different options out there and/or any links to some ressources/sample projects that you know of.
The main question I have is: Is what I am thinking possible or are there any major hurdles that I can not see yet that would prevent me from doing this? Thanks for your help