do you think the results you get from the face recognition site valid

[Seras]

They are a load up rubbish. Put my face in mostly male change my hair mostly female. Keep hair, change lighting, mostly male. Pictriev especially is bad. The other one Rekognition is inaccurate too, that one loves gendering me female and has done for a long long time mostly because I have long hair I feel.

If you want feedback on how you look submit pictures on a website such as this or elsewhere and get people to look.

[Handy]

Do not take those sites seriously; they will only provide a mild ego boost or destroy your self-esteem.

I've loaded pictures of myself into those sites and landed ALL over the spectrum. The slightest angle/off shadow/blur can dramatically impact your results.

No one has a perfect system for identifying male or female, or masculine/feminine appearance by facial features, not by a LONG shot.

[Donna Elvira]

I ran a pic of my adopted daughter through it.. 74% female. 26% male and age 34.. It lies..


That's her..

Hi Kelly,
Based on my gut feeling on seeing the image you posted I was a bit intrigued so I did a screen copy of your adopted daughter's photo and uploaded it.

Result: 95% female!
Age: 12 (OK, I assume she's a lot older than that... ?) 

It sort of begs the question, did you use the same image for your test?

Without wishing to force the issue, I suspect that if you took a sample of the avatar photos that are posted here, the results would also mostly be in line with your gut feelings... The fact of the matter is that while a "perfect" system certainly does not exist, there are more than enough characteristics of "typical" male and female faces (height width ratios, chin shape and size, forehead shape and size, eye size, size and shape of nose, upper lip height, fullness of cheeks, skin texture....) for most of us to able to instantly distinguish a male face from a female face and it can't be too difficult to codify at least some of these characteristics into a software programme.  For those that are interested, you can find a good read on the subject here: http://www.virtualffs.co.uk/My_Facial_Feminisation_Thesis.html

[Skeptoid]

Rekognition pisses me off. It'll tell me female for like nine pictures in a row then randomly decide one that is almost exactly the same as another is male. If I take it in somewhat dark lighting it's male. If I take it in bright lighting it's usually female. Except if I pull my hair back it decides male pretty much no matter what. I also tried feeding it a picture of my brother's extremely feminine wife and it said one of her pictures was male too.

Anyway, I do it to amuse myself still and stroke my ego regardless of the fact that it's a bunch of crap. The celebrity recognition most often says I look like Bryce Dallas Howard (not ->-bleeped-<-ing true at all) and Sibel Kekilli (seriously now?) so no trust from me at all.

Probably the most accurate face recognition I've tried is OKCupid.

[Zoetrope]

I think it calculates the whole thing on eyes and chin, which yeah camera angle+makeup can augment ...

It's a bit of fun 

[stephaniec]

I think it's quite fun , but I wouldn't bet the house and farm on it. I'm somewhere in between Angelina Jolie  and Jesse Jackson according to it

[ImagineKate]

I think it calculates the whole thing on eyes and chin, which yeah camera angle+makeup can augment ...

It's a bit of fun 

Other face recognition I've worked with takes faces and averages them out (an oversimplification, but that works). It appears as though pictriev is using celebrity faces to do this.