DLIB + Poisson Blending for the Mask

[ghost]

Hi guys,

I was giving a try to ofxFace Tracker and was wondering if using DLIB would not be more faster with a poisson blending of the mask ?

eg: https://matthewearl.github.io/2015/07/28/switching-eds-with-python/

and for the face landmarking with DLIB, http://blog.dlib.net/2014/08/real-time-face-pose-estimation.html

Have u ever considered to add DLIB ?

Cheers,
Luc Michalski

[genekogan]

the results from those links definitely look really impressive. i'd love to try to integrate them into ofxLearn at some point, but i haven't updated it in a while and may not free up for a little while longer. but it definitely might be time to update the version of dlib for ofxLearn, so these might be a good next step. have you been using them?

[antimodular]

check out ofxDLib for some tracking examples:
https://github.com/roymacdonald/ofxDLib

[kylemcdonald]

@ofZach has reported more robust initialization from dlib than FaceTracker. my ideal would be to have a swappable backend for ofxFaceTracker that allows you to send a flag to setup() that specifies which tracker you want to use. none of the public-facing ofxFaceTracker methods are FaceTracker-specific so it might be complex to implement but shouldn't break anyone's code.