Open ashwani-rathee opened 1 year ago
Also what's the most optimal number of images to have in img_bench? Are benchmarks going to be only about how fast a base call of some algorithm is, not caring about more qualitative analysis? Should we keep focusing on trying to find a unified structure for calling benchmarks or rather just write them? And its okay to run corner algorithms on random images?
In the long run I think the C++ interface is the right one because we can't be criticized for timing overhead due to wrappers.
Should we compare with OpenCV C++, python or OpenCV.jl? I was thinking of comparing with OpenCV python makes most sense given its what's most commonly used interface I think