Tracking multiple faces about clmtrackr HOT 9 CLOSED

Not out of the box. Theoretically it might be possible to track two faces by creating two instances of clmtrackr, and initializing them on separate bounding boxes on different parts of the screen. But you would have to fiddle with the code to get this to work properly. In practice, I think the biggest problem would be that it would run very slow, since running one instance of clmtrackr is already pretty heavy on most computers.

from clmtrackr.

hmm, isn't the face tracking using viola-jones tracking of haar-like features? meaning it will detect ALL the present faces in an image - this is the most resource hungry process by the way. then you would only have to keep track of the face models in some array. this is definitely a valid feature request

from clmtrackr.

@verybigelephants : it's very possible to detect multiple faces using viola-jones, but it's not precise enough to track facial points. The method I use in clmtrackr is unfortunately very computationally demanding, so I don't think it will work very well for multiple faces. But it's worth a try if someone wants to give it a shot!

from clmtrackr.

I just started using clmtrackr (and am very impressed btw)...but I too would be interested in multiple face functionality. I realize tracking multiples would be too resource intensive for some computers/devices...so...how hard would it be to at least detect multiple faces and expose that to the api so we could cycle through them or manually choose which one to track?

from clmtrackr.

@jb41997 do you want to just detect multiple faces? or actually track their features?

from clmtrackr.

@verybigelephants being able to detect them is all that I would really need. I can probably work around the need to track features of multiples. Tracking features of multiples would be awesome but I realize there are resource limitations, so I won't get my hopes up on that one. (are there already methods in this library that allow you to detect multiples?)

from clmtrackr.

I noticed that various functions can get passed an element and a box. For instance. Without any docs to say otherwise, it seems like the box might be a specific set of dimensions to scan within an image or video. Is that correct?

from clmtrackr.

well you could of course user part of the clmtrack but i think if just detecting face positions is the question, there are other libraries designed specifically for this task

https://github.com/jaysalvat/jquery.facedetection
https://github.com/eduardolundgren/tracking.js
(and many others i believe)

from clmtrackr.

What about such scenario:

1. user one of the existing algorithms (which is quicker than what we have here to track facial points and just track faces)
2. cut the whole scene to a small peaces with a faces
3. input those faces to the facial points tracker

Should it works effectively?

from clmtrackr.