138 comments

[ 3.2 ms ] story [ 201 ms ] thread
So this was getting slammed by reddit, and you decided to post it here as well? These poor developers.
Obligatory..."what's new?" Especially now that the site's down.

I mean, is it different from Project Oxford, the Microsoft API that's been around for awhile and is still quite amazing?

https://www.projectoxford.ai/demo/vision#Analysis

I actually tried it out early this morning, to compare it with a stock install of OpenCV 3. It got the faces correct, and the ages very well too.

Here are its guesses for the Star Wars TFA poster: http://imgur.com/XT7RmX6

Of course, perhaps users have trained it...particularly ones sympathetic to Carrie Fisher. Though I'd argue that they would've also corrected Boyega's face.

I think it's just that there is so little of Fisher's face visible. I couldn't even tell who that was until I read your full comment.
Oh dear - just put my own picture into that and it guessed I was 34 and my fiance 38...we're 25 and 26.

Edit - tried a different pic, it guessed my fiance was 51.

Well c'mon, have you really double checked? ;)

Lighting obviously plays an effect, but I was pretty surprised at how it got TFA's Han Solo down decently well. In the poster, he looks more in his 50s.

Lighting obviously plays a part. I wonder if race does as well? To use the common stereotype, does the algorithm make a guess if you're Asian, then guess downwards?

the flag for "Adult Content" reminds me of anecdotes about mechanical turk workers standing in to keep uploads child friendly, now mixed with Age Identification this gets a new perspective. Is this in active use somewhere?
Startup idea #384826 - "Does this outfit make me look fat?" As A Service
i want to invest.
Have it always says yes and then throw up an ad for FitBit. We're going to be rich.
"Free week at 24 Hour Fitness!" coupons!
I actually thought an interesting idea is an app that recommends you which dating site to use based on your face.

If you're good looking you can pretty much use whatever.

But for someone like myself who is ethnic and not visually attractive, my success rate is really low on certain sites and acceptable on certain apps.

For example, my performance on Match (Graphic I made: http://i.imgur.com/UZuSzD9.png) was pretty woeful in December. But I started using another app in the same week and had much higher success in getting responses relative to effort level.

I'm curious what worked for you, and if you have any similar data for other sites? Explaining cultural differences between dating sites/apps is something that I've tried to do multiple times, but I've never had any data to support my hypothesis.
So far I've only used Match and CoffeeMeetsBagel.

I've noticed that CMB now only matches me up with Asian women despite the fact that I'm open to dating any ethnicitiy.

I'll post my CMB stats in another time but unfortunately I can't read ethnicity/height preferences in the same fashion as I can with Match. It's a shame because it's really important data in helping me determine whether or not I should even pursue someone.

But I've bookmarked your name so that I can contact you when I gain a greater sample size for CMB and Eharmony (which I'll test soon)

I've often wondered how I'm effected. I'm half-Korean so I look Asian but I was born and raised in America and now live close to Georgia Tech. I wonder how many people assume I'm foreign.
When I was aggressively dating, I found that my success was prone to incredible fluctuations that I could never pin down. Keeping the site constant, anyway. I do suffer somewhat for being a certain flavor of queer that excludes a lot of heteronormative people.
This would be good if you can turn this into a mirror IoT device. (°ロ°) Just imagine how many people would buy this if Kim Kardashian got involved
So this is the next more computationally intense evolution of "Hot or Not"?
crowd sourcing is out and AI is in...
Today you use the crowd to train an AI...
I'd still rely on crowds directly rather than AI. I don't think AI is as accurate. You can post your photos on OkCupid and a crowd of people will rate them and help you select the best. Of course, you pay for this service by rating pictures of strangers.
Well that made me feel good. I'm way older than they suggested.
(comment deleted)
Hi, I'm Rasmus and worked on the algorithms behind faces.ethz.ch. If you have any technical questions, let me know! Sorry for having some stability issues, we got much more traffic than expected and are currently working hard to fix everything!
Your website is offline, do you have a paper about your technology?
It is back right now and they list the research.
From a data privacy point of view I would very much appreciate if you added a prominent "delete upload" button to the result page.
Yes a "delete" button would nice.
We do not store any photo on the server, so no worries!
Well, you could still add a delete button to make people feel better (I'm only half joking)
And as a potential user I'm only half-insulted.

But seriously, what purpose would it serve over a promise in TOS? A normal user will trust both (except the button would be bullshitting them). A privacy-obsessed one will trust neither.

You could make a delete button which merely opens a popup to explain that they do not save it in the first place. That way if they're looking for a delete button, they'll find it, but it will give them the right information.
"We do not save the uploaded image." from the take-this-seriously? popup.
What does it analyze?

Picture or Extracted Face?

Could this site be used to optimize your dating profile photo (if you have one)? I'm probably "Hmmm" on most photos but possibly "Ok" or even "Nice" on a few. What does the algorithm want?
Yes, it is a foreseen application.
None of the images I am uploading are working. It is saying it cannot detect a face on any of them. Is this a masking of the stability error or some other issue?
It is trying to let you know that you are not quite human.
Hi Rasmus,

Cool work! I'm curious regarding the training dataset. What is the distribution of faces by race/age? Also regarding the raters, what is their distribution? (Race/age/cultural background)?

It's widely known that attractiveness is heavily dependent on cultural upbringing.

Signed, Butthurt dev whose best pic only rated an "OK".

EDIT: You also rated Yoona (a Korean pop star) as just "Nice": http://imgur.com/uJVnQ9S. I guess that makes me feel better about my "OK". I'd stay out of Korea if I were you---I hear their fans aren't very forgiving.

Hi fatjokes,

I am Radu, one of the authors.

As mentioned in our article (http://arxiv.org/abs/1510.07867v1), the training dataset for attractiveness is from Blinq. The underage and above 37 years old face images were discarded. All people are heterosexual in our training dataset and mainly from Zurich area, Switzerland. Therefore, our model of attractiveness fits the cultural bias of Zurich, Switzerland.

We consider faces.ethz.ch a little fun tool. I hope that the fans of Yoona will understand :) With more training data from abroad Switzerland our algorithms will fit their expectations.

For age and gender we used much more and diverse data, therefore are more reliable for the majority of ethnic backgrounds.

I'm "hmmm" and 55, apparently. "Hmmm" I'll agree with, but I'm 32!
Is this a test for my attractiveness or for the AI?
Plus: The algorithm thought I was 24 years old. I'm actually 32.

Minus: I got the lowest rating possible. Haha, terribly depressing feedback before a date.

Don't feel too bad -- I ran a photo of Brad Pitt through it and he merely got a "Nice". Granted, I tried a second one and he got "Godlike". I wonder how sensitive the results are to general lighting in the photo.
I ran lichtenstein's m-m-maybe, and it returned stunning, and 26, which is probably about right, even if it's a copy of a comic.
I think ethnicity does play some role too (which maybe works with the lighting hypothesis?).

I put pictures of really attractive Asian guys (specifically men who honestly have a lot of diehard female fans who are interested in them) and at best they got "Nice"

This is actually a very interesting effect -- it exposes opinions people are generally not comfortable at expressing, for fear of being labeled as racist or something.

Sadly, asian guys are considered much less attractive than our female counterparts in Switzerland.

On a more positive note, now I know if a swiss girl likes me, she's probably not superficial. lol

Experimented more - it appears to be strongly influenced by facial expression. If I have a straight face, I'm hmm or OK. If I scowl, I'm nice. If I grin like an axe-wielding maniac, I'm stunning.

Then again, that bears up in reality. People who smile are perceived as more attractive.

I'm 41, it thinks I am 33! YAY!
It said I am 36 and I am 37. I am impressed (and Hot(tm))! However, it guessed a coworker was 34/Ummm and she is in her 50s. I am conflicted about telling her the results.
Wow! It missed my age for only 1 year (on the good side :P)!!
well, is it wrong...?
31 cat years is like 3 or 4 human years so it's probably close.

The real question is how old and hot is that popcorn?

(comment deleted)
I'm sure their algo writers are looking for new work now lol
The first processing step consist from (human) face detection. We use the standard OpenCV for our faces.ethz.ch demonstrator. A failure of this step is likely to propagate in the unreliable/wrong attractiveness prediction. For attractiveness, age, and gender prediction we start from a cropped image assumed to contain a (roughly aligned) face as found by the detector.

I hope that this helps to understand the aforementioned result.

This would make a great psychology experiment. Use an algorithm to detect someone's age, then randomly assign them an attractiveness score and see how their behavior differs based on the result. How does the random attractiveness result effect how likely they are to share their score? To retake the test? Does this vary based on the users age?
Hi sethbannon,

I am Radu, one of the authors.

We thought at similar experiments, however psychology is not our expertise. If you check our paper on hotness/attractiveness (http://www.vision.ee.ethz.ch/~timofter/publications/Rothe-ar...), I am sure that you'll find some interesting results on how different age-grouped people rate, a paradox, and more. And yes, there are many interesting experiments to do and questions to answer.

Does anyone else find it funny that the photo of Jesus didn't rank as Godlike?
Can't have been what the painter was going for...
A photo of Jesus! Is it a daguerreotype?
Finally, it's good to see we are using that AI research for the advancement of mankind. If anything this will help procreation?
Well, it was way off on my age, but it correctly gendered me ad female. Which I find quite impressive as I am a transgender woman, I've only been on hormones for 4 months, and most humans aren't even correctly gendering in me yet.

More than anything, I'm curious to know what features it was that registered me as female. Was it as simple as the long hair, or some complicated subtle mix of many small details?

Guessing gender is at worst 50% chance of getting it right. And long hair is one of the big features these algorithms learn.
How do you know it isn't guessing sex instead of gender (and hence it "sexed" you wrongly)?
> most humans aren't even correctly gendering in me yet

This made me laugh really hard. What a positive twist on the fact that the algorithm is clearly a WIP.

I've been on hormones for a little over two years now. Submitted several pictures from the last few months: it's consistently gendering me female (yay!) and a decade younger than I actually am (yay!), but it's saying that I'm ice cold "Hmm" (aw...).

What direction was it off on your age? I'm 31 (30 in the older pictures I sent), and it said I was 19-22 in all the pictures I tried.

I put Mona Lisa and she is second lowest rating. I mean, this is next door to random generator.
It possibly associates the painting texture with poor skin health.
http://i.imgur.com/Lq648My.png

Well, there goes my last ounce of self esteem.

If anyone needs me I'll be sitting in the corner with one of those criminal hacker ski masks while I work on open source stuff.

Get rid of the headphones and try again
At the time that picture was taken, I didn't own speakers.

Now, I look even worse and don't own a working webcam.

How can you look even worse? That picture is terrible, a 1 or a 2 out of 10. Anyway here's a plan for you:

Plan A) if you have any amount of disposible income (more than $0):

- spend 15 minutes reading about beginning to lift weights if you've never lifted a weight in your life

- go to the gym once (ever) i.e. on a day pass.

This plan I've listed takes 90 minutes if you start right now - as in 90 minutes from now you can be back home - and is 97% of what it takes to spend 20 minutes in a gym once or twice per week. if you've found a gym in your area and have tried it out, you're 97% of the way there. For a man, it takes basically no time whatsoever to become fit, it's literally a matter of reading about it and going there, figuring out that gyms exist. In terms of time commitment it is basically zero - in fact, because you have more energy and get things done when you're back, it's negative time commitment. You have more time in your life if you go to the gym than if you don't. Try it out and then sign up if it works. you can be back home within two hours.

Plan B) since you mention not owning speakers and not owning a working webcam, perhaps you have $0 disposible income.

- In this case begin each day with a gradually increasing number of pushups, situps, and you can do these kind of dips - https://www.google.com/search?q=chair+dip (I've listed three exercises.)

Just make it part of your day. You won't become very buff, because all of the mentioned exercises have high mechanical advantage: the reason plan A mentions the gym is because you won't find heavy weights at home and shouldn't bother trying. Home exercises aren't great. they'll get you into basic form though.

Like plan A, plan B also takes negative time (you have more time in the day - because you'll be pumped and ready to get things done - than if you didn't do it), and in terms of wall time takes maybe 90 seconds. Do it right when you wake up.

Finally, if you ever have 4 weeks where you can look scruffy and homeless, as a project you can grow a beard (simply stop shaving, and see what happens after 4 weeks), and then go to a proper barber shop to have them trim and style it for you, tell them you're good with any style and have grown a beard for the first time ever, that they should pick one for you. This will add +3 points to your look (out of 10). Don't try to pick your own style, discuss it with the barber, and those four weeks are pretty brutal, you won't like it, but there's no way to know what if any kind of facial hair you can have. Not shaving every morning takes an additional -3 minutes of your life, and this time both in wall time and projects. Going to the barber takes 30 minutes, so the net time commitment is -3 minutes times 21 days = -63 minutes and 30 minutes to find and go to a barber for their styling shave. so this, too, takes a negative amount of time.

the only people who aren't able to accept negative time for free are those who have passed away and are no longer with us. live while you can.

Of course I tried a bunch of pictures of myself, just as everybody at HN is doing, ha ha!

I didn't get above the nice level.

First picture of my wife she immediately got the godlike level. Unfair advantage!!

"Before performing any experiments we removed underage people, anyone over 37 and bi- and homosexual users as these comprise only a small minority of the dataset."

That's already biasing it of course!

"Interestingly, 44.81% of the male ratings are positive, while only 8.58% of the female ratings are positive. Due to this strong bias of ratings by women, we only predict the ratings of men."

So, the algorithm learns to rate me as a heterosexual man?

From: http://arxiv.org/pdf/1510.07867v1.pdf

Yes, this was the data we worked with for attractiveness modeling. The algorithm learns from millions of ratings from males to females and from females to males, all heterosexuals. No underage or above 37 years old, and mostly people from Zurich area, Switzerland. For age and gender prediction we use more and diverse data.

I am Radu Timofte, one of the authors.

What did you mean about the bias of the rating by women?

Is it possible that quite a few men rate every women high irrespective of looks?

Or are women just more picky? Why is it called biased?

The ratings are "like" (positive) and "dislike" (negative). If men have ratings close to uniform distribution (44.81% are "like" ratings, 55.19% are "dislike"), women with only 8.58% "like" ratings are clearly unbalanced, or if you want the women are "very picky". We call this a strong bias in the ratings of women while for men there is a small(er) bias and this only to point out the difference from the uniform distribution (50%-50%). We can not tell and we are not the ones to judge if the men are voting randomly or are less picky or if the women are right or more picky, since the attractiveness is subjective and we do not have an ultimate ground truth to compare with. We can not say if the attractiveness should follow a particular distribution, we work empirically. In our study, the ratings of men tend to agree more and correlate more with our visual representation based on the face image (looks). We removed from our data the extreme cases such as users with less than 10 ratings or with too many ratings.
It's a comb-over you insensitive bastard!
Well, if it makes you feel better, see what the algo thinks is hot…

https://polybox.ethz.ch/index.php/s/tI9QTAblYVanuA9

:-)

I've been feeding it various pictures of fursuits at different angles to see if I could force it to treat one as human, but so far I've been unsuccessful.
(comment deleted)
I've tried to get it to tell me how godlike my cat is, but it just won't.
Is that ironic? 'Cause I think polar bears are so hot.

(XPpPpP)

(comment deleted)
Feel better about yourself: http://imgur.com/uJVnQ9S
Jesus christ.

On a more serious note, training set probably didn't include data from Asia.

Edit: Just scrolled down and saw your question (and the answer you received). :C

It's only relevant if you are interested in dating a robot...
(comment deleted)
A lot of the sample photos look like they have had filters put on them. One of the things that karpathy found was that convnets were bad at images with filters.

Anyway I only got "connection error".