CrazyEngineers
CrazyEngineers
Howdy guest!
Dear guest, you must be logged-in to participate on CrazyEngineers. We would love to have you as a member of our community. Consider creating an account or login.
Replies

  • Satya Swaroop Dash

    Member • May 1, 2015

    The demo seems to be crap, I uploaded two of my photos, on the first I was 35 and on the second I was 39? I am 25 for crying out loud.
    Are you sure? This action cannot be undone.
    Cancel

  • rahul69

    Member • May 2, 2015

    For me it gave 34, guess this software is quite inaccurate, the results would have been better if there was an option of user feedback of some kind so as to improve results using machine learning.
    ash
    Microsoft's machine learning team built a ....
    Of course, seeing we are crazy engineers here, let's discuss how you think the algorithm works? Goes beyond your typical facial recognition engine.
    Well looks like u missed the link on "Story behind the demo", the link is on that page (howold.net). There they have explained how they created it,
    Well if I have to create this, apart from face recognition engine, I will set certain parameters which will help in better diagnosis, eg: facial texture eg. if smooth on cheekbones (of course I am being wildly generic 😁), looking for areas where wrinkles are present. I know my Ideas needs refining, but I guess these would be a good start, what do u think ?? 😀
    Are you sure? This action cannot be undone.
    Cancel

  • Ankita Katdare

    Administrator • May 2, 2015

    ☕ Mine shows 37. 👀 Do I need to say that it's not even remotely close.

    (Looks like never using any cosmetic products has taken its toll 😁)

    They wrote there that they are improving their API. I took the test only because it's becoming so popular and found that most comments on it are positive. A quick overview said that most people who say that their age was guessed correctly are white. If all the 3 of us above got it wrong, has it got anything to do with skin colour?

    I am also thinking that the lighting conditions affect the output a lot. My photo was in bright sunlight.
    Are you sure? This action cannot be undone.
    Cancel

  • Ashraf HZ

    Member • May 3, 2015

    Isn't it good to look like in your 30's? 😁

    rahul69
    Well looks like u missed the link on "Story behind the demo", the link is on that page (howold.net). There they have explained how they created it,
    Well if I have to create this, apart from face recognition engine, I will set certain parameters which will help in better diagnosis, eg: facial texture eg. if smooth on cheekbones (of course I am being wildly generic 😁), looking for areas where wrinkles are present. I know my Ideas needs refining, but I guess these would be a good start, what do u think ?? 😀
    Yep, checked the link. Unfortunately it doesn't give the grunt details of the Face API in terms of estimating the age attribute (I guess it's their secret sauce). Good idea on facial textures.

    Can't seem to tag #-Link-Snipped-#. I remember he was working on some MS based research projects. Wonder if he still lurking on CE these days.
    Are you sure? This action cannot be undone.
    Cancel

  • MaRo

    Member • May 3, 2015

    Hello Ash, Hi CEans!
    I only had some talks with Microsoft about my project, I've no relation to Microsoft 😁
    About this Age guessing, I think it's technically great, of course the results are mess, but it really can capture your face, like when I tried to upload a photo for myself where I look good (shaven beard, not tired face...etc) It got me 23, but with another photo with not very good looks, it got me 33, I think these results are great step in the way of "Age recognition" instead of "Age guessing", of course I've to test on like 100 photos to be sure, but no time for that! haha 😁
    Are you sure? This action cannot be undone.
    Cancel

  • Ashraf HZ

    Member • May 3, 2015

    Yo MaRo, welcome back! What are you up to these days man? Any new cool projects? 😀
    Are you sure? This action cannot be undone.
    Cancel
Home Channels Search Login Register