is there a way to enable flashlight/flash on regular camera

asked 2016-09-08 06:58:43 -0600

atv gravatar image

Is there a way for me to enable continuous flash or a flashlight/torch mode if you will as on mobile devices, for the isight camera (or any webcam).

I've seen examples on how to do this with android (and i'm sure not all devices support it), but as my isight camera supports at least flash, i should be able to turn it on all the time ?

Thanks

edit retag flag offensive close merge delete

Comments

not an opencv question.

berak gravatar imageberak ( 2016-09-08 23:16:05 -0600 )edit

Ok. I thought it was because i came across similar questions but for mobile platforms which do indicate opencv can do this: http://answers.opencv.org/question/26...

atv gravatar imageatv ( 2016-09-09 03:14:44 -0600 )edit

apologies, then, i was wrong !

(still, i dont think, it's possible)

berak gravatar imageberak ( 2016-09-09 04:50:30 -0600 )edit

Yeah i'm not sure about it either. The reason i ask was because i was trying to find ways to illuminate the scene. I came accross some articles which say that autoexposure on iSight cameras on imacs (dont know about other webcams) is messing with face recognition and scene light stability. But that might be material for a different question.

atv gravatar imageatv ( 2016-09-09 04:55:04 -0600 )edit

indeed, anything auto (whitebalance,exposure, focus) is a pain with computer vision.

so you wanted to overcome that by turning the light on (control the lighting) ? makes some sense.

berak gravatar imageberak ( 2016-09-09 04:59:36 -0600 )edit

I guess. I'm just reaching out really, trying different things. Was reading about NIR (near infra red) options too which apparently completely eliminates illumination problems.

For example, my code works perfectly in the morning, but in the evening all of a sudden it doesn't. Lighting you would say. But my code tells me the lighting is nearly the same (it measures the lighting in the room). So it must be more complicated then that (indeed, autoexposure, exposure, focus, local shadows, global shadows etc).

In my amateur mind i had hoped to find a optimal lighting range for facial recognition, and then modify the actual lighting range by adjusting the brightness, blur or contrast accordingly to get to that optimal range again.

But as said, it doesn't seem to be that easy.

atv gravatar imageatv ( 2016-09-09 05:08:14 -0600 )edit
1

actually, bioinspired::retina worked wonders for me there (software lighting normalization)

berak gravatar imageberak ( 2016-09-09 05:33:55 -0600 )edit

Hey, thanks for that. I never saw that before. I'll try out the webcam demo tonight. So you just let this do it's work before you pass it's to face recognition? Do you use parvo or magno?

atv gravatar imageatv ( 2016-09-09 06:36:59 -0600 )edit

magno for this. parvo is some kind of saliency map, interesting, too !

remember, to apply this kind of preprocessing to both train & test images !

berak gravatar imageberak ( 2016-09-09 08:35:32 -0600 )edit

Gotcha, thanks for the heads up. Did you get much improved recognition results with this? I will try this out tonight.

atv gravatar imageatv ( 2016-09-09 08:49:43 -0600 )edit