Before I run into issues way down the road I want to ask my question on here.
When storing descriptors by using the GPU enabled functions surf() or orb() are the descriptors stored per row of a GpuMat? I have seen the documentation state that the CPU based versions of the functions store descriptors per row of a Mat datatype and I want to be sure that the GPU version does the same. I'm not sure because of the use of the "rowrange" function to set a range of rows to pass to the kernels to calculate the descriptors. It makes me think that maybe the descriptors are across several rows but I can't yet tell by just looking at the code. I figured I may get a quicker answer on here.
I just want to store each descriptor individually from the GPU feature detectors to put into a float array in my own data structures.
Any Feedback is appreciated.