WebGL only? about clmtrackr HOT 10 CLOSED

It should work on browsers without WebGL as well, but the tracking might be a bit slower.

from clmtrackr.

@auduno ok thanks :) i am testing the lib on cordova / phonegap , i notice one thing:

iOS safari v.8 detected all the faces in pics
while android 4.1 browser do not detect the same faces in the same pics.

I don't have errors, so i tought it was a matter of webGL, since android doesn't support webGL.

Anyway, thank you i will try to debug better :)

from clmtrackr.

@auduno Hi, sorry for interrupting again, i just copied your example (http://auduno.github.io/clmtrackr/clm_image.html) and all the scripts, i find hard to understand which scripts are needed and which not, actually i just need to recognize faces in photos whitout watermarking/overlaying or modify on the photo.

Just to be sure, can you just tell me which of these files aren't needed for the scope:

utils.js

numeric-1.2.6.min.js

mosse.js

jsfeat-min.js

frontalface.js

jsfeat_detect.js

left_eye_filter.js

right_eye_filter.js

nose_filter.js

model_pca_20_svm.js

clm.js

svmfilter_webgl.js

svmfilter_fft.js

mossefilter.js

Stats.js

Thank you and sorry again

from clmtrackr.

If you're just looking to detect faces, you should actually just use jsfeat (which is not my library). I've created a wrapper for jsfeat, jsfeat_detect.js, which makes it a bit simpler to use, so you could take a look at that. So you'd just need the files jsfeat-min.js and jsfeat_detect.js.

from clmtrackr.

@auduno thank you a lot 👍

from clmtrackr.

@auduno Sorry i am here again, i wanted just to ask you one last thing if possible, i included the jsfeat.min.js and the jsfeat_detect.js so i did:

var f = new jsfeat_detect(document.getElementById('canvas'));

f.findFace();

console.log(f);

It works and i get a returned Object in console, which contains data like (confidence, x, y etc..), but i don't understand how to check this data to match a face, may i ask you how you did check this?

Thank you and sorry again

from clmtrackr.

The position of the detected face is given by a box with coordinate of top left corner at (x, y) and coordinate of bottom right corner at (x+width, y+height), where the data x, y, width and height is given by the returned object. The returned data "confidence" tells us a score of how sure we are that there is a face in the bounding box, higher scores are better.

Note that in jsfeat_detect.js I only return one face even if there are many faces in the image, I only choose the bounding box with highest confidence score.

from clmtrackr.

@auduno clear now thank you! Sorry for bothering you

from clmtrackr.

@auduno Hi,

Can i use this for other object detection like hand and how?

from clmtrackr.

Yes, probably, but you'll have to train your own model of the hand using for instance clmtools. For tracking hands, I think it's better to take a look at js-objectdetect since it contains a ready-to-go model of hand tracking.

from clmtrackr.