Using code in video processing,how can I speed up it?

[AiNi-KuangBiao]

Dear uricamic,
I am a beginner of OpenCV . I want to use this code in real time video processing,but i find this project is no as fast as previous flandmark project .Function cvImgToCImg and detectMultiScale in video_inpuit.cpp take a long time. How can I achieve the speed show in page http://cmp.felk.cvut.cz/~uricamic/clandmark/.
Sorry for my poor english.
Many thanks.

[uricamic]

Dear @AiNi-KuangBiao,

I am not aware of such problems. Actually, the newer version is even faster (if the featurePool is used), compared to the old library flandmark, there was a significant speedup (from 10ms to 2ms on the 8 landmarks model).

Just a quick check, do you compile and use the release configuration?

Please show me some screenshot, with your processing times, or describe the process how you figured out that cvImgToCImg function is the bottleneck. Btw, if really this function is a bottleneck, you can switch to some different different image processing library than OpenCV, CLandmark core is completely independent of it, it is used just in C++ examples...

[AiNi-KuangBiao]

Dear @uricamic ,
I am sorry,I compile it in debug model. I will try out it in release model again.
Thank you very much for your reply.