Python version not as robust as CPP version

[soulslicer]

I just ran some of my own datasets on the CPP version. It worked great, but it was very noisy (keypoints detected all over) in the python version. Why might that be?

[noodlebreak]

Can you give some images from your differing dataset images?

[dtbaker]

Anything further on this @soulslicer ? I just got up and running with the Python version on a Raspberry Pi, contemplating digging into CPP if it is going to run better.

[soulslicer]

I ended up not using CMT as I found particle filters used together with classifiers to be more robust.

[dtbaker]

Thanks for the reply @soulslicer

This is where I am at now with CMT object tracking:

https://www.youtube.com/watch?v=hvzohGuw8XI

https://github.com/dtbaker/pi-web/blob/master/CMT/start.py

This is my very first attempt at OpenCV and Python so it's a little rough :) I hate asking for help on these sort of things, but do you have any links to tutorials about particle filters/classifiers? There just seems to be so many options out there. Need someone to tell me "use this and this" for best tracking results :)

Cheers

[duzefu]

probabaly because CppMT use FAST detector but CMT use BRISK detector .FAST detector can find almost 10 times keypoint than BRISK