Computer Vision News - November 2016

Thus, I would like to recommend two things: the first one is to start this migration very early in the project and replace the Matlab function with MEX (an envelope linking between Matlab and C++) and thereby use the equivalent function in C++ : you still enjoy the flexibility of Matlab and at the same time at a lower level use the C++ implementations which will later be used in the C++ final version. In this way, the gap between the two environments will be much smaller . After that, even when you implement several projects of several months each, there is no need to wait for the conclusion of the research; and as soon as confidence with some of the steps is reached, the porting can start; for example, the edge detection of canny with its parallel in OpenCV. In this way, even when there are some changes, they can be handled and the software can be tweaked as proper. My second recommendation is to get into details on the differences between the functions . For example, the canny edge detector in Matlab is much more sophisticated and it offers automatic adjustment of the parameters, while OpenCV is limited to more basic parameters and it needs to be tweaked manually. When one is prepared to this difference, it becomes possible to predict the required level of effort and resources needed to port all functions from one environment to the other. “ Start this migration very early in the project and replace the Matlab function with MEX ” Management Computer Vision News Project Management Tip 71