Create a Robot Image from any text string

Post by e1ven » Fri, Jul 08, 2011, 2:37 PM UTC

MattBearman wrote: post #9

I love this, simply because http://robohash.org/mattbearman kinda looks like me :D My one criticism is that it took me while to figure out (and I'm web dev) It wasn't clear until quite far down the page that you need to just put the text string after the URL. I'd recommend having a text box into which pasted text can be robohashed prominent on the homepage, as well as clearer instructions. Edit: just noticed there is…

Post by andrewcooke » Fri, Jul 08, 2011, 2:39 PM UTC

e1ven wrote: post #10

andrewcooke wrote: post #5

this is cute. if anyone is looking for a more abstract approach i threw some code together a year ago that generates colourful "mosaics" - http://www.acooke.org/hash-icons.html (i really need to improve the page and release the code...)

I love the mosiacs; They remind me of a prettier version of identicons. My concern was just that people won't remember a random pattern of dots very well.

Post by e1ven » Fri, Jul 08, 2011, 2:44 PM UTC

praptak wrote: post #14

It's similar to the StackOverflow April Fools' Unicorn Avatars: http://meta.stackoverflow.com/questions/37328/my-godits-full... Which you can try out at: http://unicornify.appspot.com/use-it

Post by EricBurnett » Fri, Jul 08, 2011, 2:48 PM UTC

e1ven wrote: post #7

pavel_lishin wrote: post #4

Earlier quoted context omitted.

What are the odds of a collision? How many unique elements are there that can make up a robot?

If you use all three sets, and enable backgrounds, the collision chance is much lower. Each Robot is made up of 6 sets of 10 choices each, plus backgrounds. It's not perfect, but it's enough to give you a basic view into "Is this the same guy or not"

Post by e1ven » Fri, Jul 08, 2011, 2:55 PM UTC

EricBurnett wrote: post #17

e1ven wrote: post #7

Earlier quoted context omitted.

If you use all three sets, and enable backgrounds, the collision chance is much lower. Each Robot is made up of 6 sets of 10 choices each, plus backgrounds. It's not perfect, but it's enough to give you a basic view into "Is this the same guy or not"

as in 10^6 possibilities, ignoring backgrounds? For the curious, that means collisions are expected >50% of the time with ~1200 samples, and someone could easily generate collisions on purpose with little effort. So it works well for a simple signal (which is the goal), but shouldn't be relied on if people might try to abuse it.