Show HN: Facerank — A/B test your Tinder pics. What do you think? • Hacker News Board

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#11

Post by ksaitor » Fri, Feb 22, 2019, 12:52 AM UTC

verdverm wrote: post #6

How about generating better ones using deep fake tech?

lol. great idea — and then sending them back to the owner?

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#12

Post by ksaitor » Fri, Feb 22, 2019, 1:06 AM UTC

charlieegan3 wrote: post #5

My feeling is that you need to let users choose the demographic rating their photos. This clearly has a bit of a cold start issue though. I'd really love for a service using ELO/relative ranking to exist and exactly for this purpose. I've used https://www.photofeeler.com in the past - I think you're looking to give users a similar service. I have greater faith in your ELO method though. Perhaps to give you some motiv…

Thanks for sharing! I've seen Photofeeler.com around.

Have not added ELO based rankings, but might be fun to implement this in future.

What and how do you think I should change for in Facerank?

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#13

Post by lapnitnelav » Fri, Feb 22, 2019, 1:46 AM UTC

What's in it for the people doing the rating?

Follow up question : Given the target audience (dudes mostly), why should they help you rather than purposely misguide you into choosing the worst ones in order to undercut competition? A bit twisted I know but hey, if I thought of it, I'm probably not the only one ;)

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#14

Post by ksaitor » Fri, Feb 22, 2019, 2:21 AM UTC

lapnitnelav wrote: post #13

What's in it for the people doing the rating? Follow up question : Given the target audience (dudes mostly), why should they help you rather than purposely misguide you into choosing the worst ones in order to undercut competition? A bit twisted I know but hey, if I thought of it, I'm probably not the only one ;)

What's the motivation? Give it a try… it's fun to vote.

I've posted this on Product Hunt and Reddit this week, and so far people did 33426 votes on pictures. Pretty crazy!

Interestingly, a few comments (yours and on Reddit) assume how "bad" people can be. Surprisingly, so far, overall voting behaviour been very positive. People have not been abusing votes, comments nor images in the past few days.

So far 170 users uploaded 218 photo pairs - and all of them are genuinely good.

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#15

Post by ksaitor » Fri, Feb 22, 2019, 6:08 AM UTC

verdverm wrote: post #6

How about generating better ones using deep fake tech?

Actually, gave this a second thought…

Might be cool, if done right:

   - upload a few profile pics
   - GAN generates an optimized pretty profile pic for you,
     that you can used on Tinder and other social media

I'm pretty sure people would be willing to pay a couple of bucks for that.

How hard would that be to train a model like that?

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#16

Post by charlieegan3 » Fri, Feb 22, 2019, 7:05 AM UTC

ksaitor wrote: post #12

charlieegan3 wrote: post #5

My feeling is that you need to let users choose the demographic rating their photos. This clearly has a bit of a cold start issue though. I'd really love for a service using ELO/relative ranking to exist and exactly for this purpose. I've used https://www.photofeeler.com in the past - I think you're looking to give users a similar service. I have greater faith in your ELO method though. Perhaps to give you some motiv…

Thanks for sharing! I've seen Photofeeler.com around. Have not added ELO based rankings, but might be fun to implement this in future. What and how do you think I should change for in Facerank?

I think being able to say, "only allow women/men aged x to y to rate my photo" would be the main one.

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#17

Post by verdverm » Fri, Feb 22, 2019, 10:32 PM UTC

ksaitor wrote: post #15

verdverm wrote: post #6

How about generating better ones using deep fake tech?

Actually, gave this a second thought… Might be cool, if done right: - upload a few profile pics - GAN generates an optimized pretty profile pic for you, that you can used on Tinder and other social media I'm pretty sure people would be willing to pay a couple of bucks for that. How hard would that be to train a model like that?

Yeah, I bet they would given they are willing to pay for super likes. I could definitely see using it for all sorts of reasons, like automated image brushing.

I do need a new LinkedIn head shot...

You could also get into generating stock photos for websites and other media like marketing

Re: Show HN: Facerank — A/B test your Tinder pics. What do you think?

#18

Post by coralreef » Sat, Feb 23, 2019, 2:30 PM UTC

ksaitor wrote: post #4

mtmail wrote: post #3

There always was a market for such a service "Hot or Not was preceded by the rating sites, like RateMyFace, which was registered a year earlier in the summer of 1999, and AmIHot.com, which was registered in January 2000 by MIT freshman Daniel Roy." https://en.wikipedia.org/wiki/Hot_or_Not "Hot or Not was sold for a rumored $20 million on February 8, 2008 to Avid Life Media"

True true! There is a bit of a difference tho. "Hot or Not" was used to compare two different people. My inspiration for Facerank was to help me optimise my Tinder profile, lift by ELO score and get more matches (maybe a wife, lol). So here, people supposed to upload their own pics :)

I don't recall HotOrNot using people comparison, but a 1-10 hotness scale.

FaceMash compared people (at least according to the movie)