Hi, I was wondering if any thought has been given to bringing this implementation to OpenCV? As of now, OpenCV lacks a CUDA implementation of SIFT and this work seems to be the fastest implementation that remains faithful to the original SIFT algorithm.