AI Tailors Can Wait (bloomberg.com) 71
Bloomberg Businessweek: Original Stitch has all the trappings of an e-commerce success story. The pitch is simple: Original Stitch uses computer-vision software to review photos of your most beloved dress shirts uploaded to the company website, then delivers perfectly tailored copies. We tried it -- the only problem was that it didn't work. When the first shirt arrived too tight around the chest and too long in the sleeves, we figured an editor's sloppy photography was to blame, but the problems persisted with a second attempt. A third shirt, ordered under a different name to make sure we wouldn't get special treatment, could barely be buttoned up. The sleeves felt like tourniquets. "We tried to push the envelope," Original Stitch founder Jin Koh acknowledged after we confronted him with the results. "Obviously, it's still in beta." In December, three months after launching the service, Koh quietly pulled it down. He's returned to asking users to fill out a questionnaire with their own measurements while he works out the bugs.
Re: (Score:2)
Why test at all? (Score:3)
Why test your product at all if customers can just do it for you?
1. Release product under name #1
2. Get a lot of test data from thousands of soon to be disgruntled customers
3. Close up shop
4. Release product under name #2
5. Profit
Denise (Score:3)
I say we name this AI, Denise [pinimg.com].
Sure the AI has some flaws now, but it will eventually get better. Also, a lot of the problems are going to come from the quality of the picture. The lens usually distorts the image. I remember doing an AI project one and we had to account for the fish-eye effect of the lens. Also, our adjustments only worked for our camera and would have been broken for another camera.
Re: (Score:3)
I'd say Keystone Effect will have a bigger influence on distorted garments than any lens distortions.
If you take a look at the Bodygram [originalstitch.com] page at Original Stitch the example photos are taken from directly above the shirt.
The first picture in TFA demonstrates the Keystone Effect well - instead of the camera being directly above the shirt when the picture was taken it was off-center by a couple of feet and taken from below. This resulted in the top of the shirt appearing about 70-80% of the width of the bottom
Re: (Score:2)
Sure the AI has some flaws now, but it will eventually get better. Also, a lot of the problems are going to come from the quality of the picture. The lens usually distorts the image. I remember doing an AI project one and we had to account for the fish-eye effect of the lens. Also, our adjustments only worked for our camera and would have been broken for another camera.
That's the least of their worries, even if the system worked perfectly human tailors have little to worry about. The reason you go to a tailor these days is for the service. A proper tailor will do 2 or 3 fittings at a minimum as nothing ever comes out perfect. A made to measure tailor let alone a completely bespoke one will have you try the clothing on to determine if it needs adjustment to your body shape and it usually does.
The AI tailor will be competing with off the rack stores for business.
It isn't AI (Score:2)
Re:It isn't AI (Score:5, Interesting)
http://www.vetta.org/documents... [vetta.org]
which leads to...
https://en.wikipedia.org/wiki/... [wikipedia.org]
It's a form of A.I.
It's not strong A.I.
Strong A.I. could be an extinction level event for humans. I think limits on available power will slow it down enough for us to have time to react. But there's a significant chance for a failure of friendliness combined with superhuman intelligence and superman manipulativeness. And people researching strong A.I. don't appear to be taking sufficient safeguards to me.
Re: (Score:1, Flamebait)
Re: (Score:2)
Re: (Score:2)
Re:haha (Score:4, Insightful)
Subjective is right. I got some made to measure shirts in Taiwan which fit like a glove and were hence unwearable. I ended up getting them to clone my favourite shirt which was quite a bit looser and thus wearable with a suit for eight hours in a place which is usually boiling ass hot.
Meanwhile it's not all that hard to buy ready to wear shirts which fit OK.
The problem is not creating clothes that fit like a glove, it's the subjective deviations from that which everyone is used to.
Re: (Score:2)
I've worn stuff like that, I don't get guys that wear even tshirts that tight. It's like being saran wrapped in sandpaper once you start sweating.
Re: (Score:2)
As someone who is just under 6' and just around 260lbs, I need a little extra room
With a BMI of 36+, you meet the classification for Severely Obese, and need a big extra room.
Re: (Score:1)
Yes, I'm a big guy, I'm aware of this.
But BMI is not a great measurement without other stuff like your actual muscle mass or percent body fat.
I'm pushing 50, so I'm not as fast as I used to be, but I can still lift far more than most people, and still knock most people on their asses. Based solely on height and weight, I've known people with little body fat who have been deemed obese -- they'd have to cut off a lim
Re: (Score:2)
You are ignoring the chicken-or-the-egg problem. Most men don't care about fit, but it could be because most men have never had clothes that fit, outside of maybe prom and weddings.
The sane strategy for AI tailors is going to be reaching that "good enough" market first, and creeping upward as the tech improves and better data is collected.
Re: (Score:2)
When I wore clothes that other people thought fit, I was quite uncomfortable. I like a loose fit, I hate ties, etc.
OTOH, this *is* highly dependent on the material. a good stretchy material that breathes well can fit as closely as it want...around most of my body. But not around my neck. So thermal underwear (in season) is comfortable. But when I bend, it stretches to fit. That makes a big difference. Something equally close fitting that confines movement is just unacceptable. Hell, even something f
Re: (Score:2)
Re: (Score:1)
AI (Score:1)
Bumper Sticker Quality (Score:5, Funny)
"We tried to push the envelope, ... Obviously, it's still in beta."
This needs to be printed on a t-shirt. I would wear it everywhere.
wrong blame (Score:2)
It sounds like it's the data-gathering mechanism that is insufficient to the task, not that "AI tailors" are to blame.
It's pretty flipping hard for anyone to take a picture (much less something sloppily done) and make a wearable, nicely-fitted garment from it. There's some pretty complex 3d geometry going on.
Re: (Score:2)
Yes, you'd need at least two photos taken at right angles to give a good image, and that probably wouldn't suffice.
OTOH, a post above claimed that camera lenses systematically distort things sufficiently that that wouldn't even come close. The the processing needs to be tailored to fit the camera that took the photo. Whee! That sounds like it could be fixed by taking the photos standing against a grid. This is getting more and more complex. Now how do the customers get the standard grid, and how do you
Obvious solution is obvious (Score:2, Interesting)
The obvious thing here is that they are trying to copy an existing thing, not by taking a picture of the wearer, but by taking a picture of the item. Take a look at all the chinese counterfeit websites (here's one: gamiss.com ) generally any website that is 85% off and looks like a generic shop with a label applied and uses the same photos you see on eBay are selling counterfeits (anything "new without tags" is counterfeit.) Now take a look at the reviews for these sites, you will see the same thing
"Ill fit
I will buy from an online tailor named... (Score:3)
Re: (Score:2)
I would only buy from an online tailor called Pendel and Braithwaite. "Tailor of Panama" reference. Great movie.
Re: (Score:2)
I will buy from an online tailor named Garek. That is all.
Is he off Keeping up with the Kardaisans?
You have been lucky! (Score:2)
"Most beloved dress shirts?" Really? (Score:2)
Is "most beloved dress shirts" something normal people actually have? Sounds absolutely ridiculous. Using computer vision to determine measurements would be great, but you don't need to clone an existing shirt. That's just silly. Are these even going to be cheaper if they're custom manufactured for each user?
Seriously dumb implementation ruins great idea (Score:2)
So instead of taking actual measurements, that require only maybe 5 minutes of effort, your try to do a shirt off of a photograph only? How is that a good idea, or even reasonably chanced at success? People come in all different shapes and sizes, and I fully expect to have 100% high quality custom tailored clothing within the next 10 years courtesy of robotics and AI, why make it harder and less likely to succeed than it has to be? Nail down quick and efficient custom shirt AI fabrication and have kiosks