[DaveMorton]

Ok, this isn't directly AI related, but I know of some individuals who may find this interesting:

After struggling with several different types of CAPTCHA image scripts out there (recognizing them, not using them) I decided to write my own a while back (which you can see on my website's Contact Page), which, unfortunately, also gives me trouble, though it's nowhere near as bad as others that I've tried.

Anyway, I took another look at some other scripts that have been developed over the years, and found the notion of using image recognition to be rather interesting, so I decided to do a "proof of concept", and see how it went. The result is at http://www.geekcavecreations.com/captcha_test/captcha.php and I'd like to get some feedback on it. Bear in mind that the current script only deals with a small number of randomly selected images, so there's a lot of repitition, but there should be enough to give you an idea of what to expect.

[Bragi]

It's a lot easier to answer (I find some captchas unreadable).

[DaveMorton]

...I find some captchas unreadable...

So do I, Jan. Even the earlier one that I wrote. This one is geared more toward ease of 'human readability', but I'm not sure how easy/hard it would be for a computer script to suss out it's meaning. I'm figuring that a good OCR script could likely figure out the question, but answering it correctly would be another matter, I think.

BTW, did you refresh the image a few times. to see a good variety of it's possibilities?

[Bragi]

with sufficient energy, all captchas can be broken (for instance, you can redirect them to another site and have real people answer it for you). I think, the real hurdle here is to have enough variation. Most captchas are auto generated, so they can have more variance. With the image-question system, it's a bit harder and more work perhaps.

[Data]

Anything that helps make these captchas easier to read for humans has to be a good thing.

One of the problems I have with them is my colour blindness, on your first post in this thread GCC I couldn’t say what colour the cat is, maybe yellow or orange I couldn’t say, but I could say he’s a nice cat 


EDIT:

My other half has just told me that the cat was green, I completely and utterly failed   

[DaveMorton]

Colour blindness can be an issue, but that's why less than 30% of the CAPTCHA questions deal with colour. And, of course, I know that the age old maxim "locks are only for the honest folks" is just as true here as it's original intent, but if helps keep me from wading through seas of 'male enhancement' ads, then it's worth the trouble.

[Kappers]

I think that's a brilliant alternative to the current irritating-as-hell system. I have an idea that may be moving onto a tangent slightly, yet it will avoid any form of color issue, and will finish with a very simple integer answer! How about you bring up a randomized shape, and ask them to multiply the amount of vertices by a randomly generated number? I'm not so knowledgeable about captcha yet it seems reasonable to me, except potential limitations of randomization.

[sunama]

It's a lot easier to answer (I find some captchas unreadable).

I cannot describe how much I HATE captchas. I find most of them unreadable.

[sunama]

OP...Are you seeking to turn this into a commercial project?
I urge you to try and market this idea, because your 'captch alternative' is 1000x better than the current captcha.
Really, all you need to do is come up with new questions and then start marketing the idea. I doubt you would even need funding. You could start earning ALOT of money doing this.

[DaveMorton]

@Kappers:
I'm sorry I missed your post when you made it. I'm not so certain about asking a user to use math skills that may or may not exist. I'm trying to keep this as simple for the user as possible while still confounding the bots. I'll consider your idea, though. It may have merit.

@Sunama:
The possibility of making this a "commercial" venture is certainly worth investigating, but as you've pointed out, it will need (at the very least) more questions to ask. It will also require all new artwork, as what I'm using now is not anything that I've created (well, at least not the cartoons), and thus I don't hold copyright privileges. In order to keep the script package small, any new images will have to VERY simple, and have a low color count, just like the images I'm currently using. Unfortunately, I can't draw to save my life. I can certainly manipulate images well enough, but to create one that's "hand drawn" is beyond me.

[Freddy]

They don't need to be works of art Dave I think even the simplest line drawings would be fine.

[DaveMorton]

I'm looking at a site now that offers free clipart that will suit my purposes. And the nice thing is, they're quite numerous and varied, and simple enough to do what I want with them. Now if I only had an SVG editor. (there's also an entire section on SVG images that would go nicely with this)

[sunama]

OP. What you need to do is goto some clipart OR artist website.
Get yourself some contact details of some artisits.
You then pitch to them and ask them to work with you, to come up with some drawings, as per your requirements.
He will do the artwork and you will do the technical/marketing work.
You give him a 20% split of any earnings from the project.
This is not difficult and I am pretty sure you can find some artist who can work with you and do the drawings in his spare time.

Please don't use the excuse of being unable to draw as your reason to not advance your idea.

The best thing about your idea is that it will require virtually no investment.

[DaveMorton]

Actually, with this new resource that I'm investigating, I may not have to worry about collaboration at all. And if I can integrate SVG processing with my script (something I have yet to look into), then I may not have to even worry about having images at all (so to speak), since SVG is a form of XML, and PHP processes XML quite handily. But for now, I'll just stick to images, of which I now have a considerable number.

Oh, and my friends call me Dave.

[sunama]

If you are smart, you will be able to launch your idea into a commercial venture, very quickly. The reason is that your idea is very straight forward. Your main problem is creating new questions/images (which is time consuming).

What you could do is create a website especially for your idea and then start emailing different companies and forums which currently use captcha, to see if they are willing to change from the current captcha to your idea.
Once you have some ready and willing "customers", you can start increasing the number of images/questions.
If you do this correctly, within 30 days, you should be ready to launch.

Another idea is create a questionnaire and post it up on Facebook (or similar). Ask people to fill it out. You can create an advert. Basically, you want to prove (by market research) that people prefer your idea over captcha.
Give your idea a brand name, asap and start marketing it. Create a brand. It is the brand which is going to be worth the money.

Keep things simple. Simple. Simple. Simple. Some ideas become so complicated that they implode and die before even starting. Some people on this very forum are doing just that (I'm not naming any names).
Launch things 1 step at a time. Don't try and conquer the World in 1 step. Break it down into multiple smaller steps.

The first version of your software is supposed to be poor. Your algorithms must evolve with time (this is important). Your fundamental idea though, is better than the current market leader.

Anyway dude, good luck and if you need people to fill out questionnaires, post the links on this website.