ORB theory and implementation questions

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

ORB theory and implementation questions

opencv-users mailing list
I am experimenting ORB to be used in FPGA. Before that, I need to test and improve ORB computational cost which require me to understand every bit of ORB theory.

1) I was reading a few papers which mention about centroid. Anyone have any maths proof on how this works on orientation compensation in ORB ?

2) Could anyone elaborate more on the hardcoded sampling pairs generated in https://github.com/Itseez/opencv/blob/master/modules/features2d/src/orb.cpp#L375 https://github.com/Itseez/opencv/blob/master/modules/features2d/src/orb.cpp#L375 with the method as mentioned in https://gilscvblog.com/2013/10/04/a-tutorial-on-binary-descriptors-part-3-the-orb-descriptor/ https://gilscvblog.com/2013/10/04/a-tutorial-on-binary-descriptors-part-3-the-orb-descriptor/  ?

3) If I want to improve computational cost of ORB, where shall I focus on ? I have been reading literature papers for few days already. The code I am using are at https://github.com/Itseez/opencv/blob/master/modules/features2d/src/orb.cpp https://github.com/Itseez/opencv/blob/master/modules/features2d/src/orb.cpp , https://github.com/Itseez/opencv/blob/master/modules/features2d/src/feature2d.cpp https://github.com/Itseez/opencv/blob/master/modules/features2d/src/feature2d.cpp , and http://pastebin.com/N69WE89j http://pastebin.com/N69WE89j

4) In my final step, I need to use it in FPGA. Anyone have experience doing openCV C++ processing in FPGA (RTL) ? Please guide.

Thanks !