Closed paulSerre closed 3 years ago
Hi,
I did with ImageCapture.cpp with an img_reader, but I have a really low fps and I want to do real time... Do we get a better fps with video?
Okay. I figured out that sequence_reader keeps in the CLNF class the information of the previous pose, which allow to save time for the following landmark detection in the next frame. That's why I hadn't a good fps. I fixed it.
Hello,
I made a server of FeatureExtraction (only for Gaze, HeadPose and EyeLmk): it receives a video stream from a front-end via a websocket. I would like to know which is the best utility to use (SequenceReader, ImageCapture) ?
I think I have to use "SequenceCapture.cpp", add a function to open a socket instead of webcam, and do aproximately the same thing as with the webcam, but I want to be sure that by doing this, I won't have a problem of incoming fps higher than outgoing fps because I want a real time system.