performance between ORB and BRIEF

[appleleaves]

If we turn off the scale/rotation invariance functionalities of ORB and BRIEF, these two descriptors should be the same. However, it shows different performance in the paper. Why?

[vbalnt]

Hi @appleleaves

Can you elaborate on why they should be the same? The two descriptors have completely different sampling pattern, BRIEF is completely random, while ORB is optimised and structured.