“Do I look fat in this?” … “You do now!” (Kinect FAIL)

Here’s an excellent idea: use Kinect to display clothes on people in real-time, inside a fashion retail shop:

http://ar-door.com/2011/05/virtualnaya-primerochnaya-dlya-topshop/?lang=en (Scrub to 0:25 – first 30 seconds is moronic marketing-person bumf)

This has huge potential:

  1. Much much faster than browsing – potential for more positive purchase decisions in less time
  2. Less space lost to changing rooms – fewer rooms needed for same number of customers (space is often at a premium in retail outlets)
  3. Show an enormous range of stock while keeping very little on-site
  4. Show every size, rather than just the sizes that are in stock on-site

But hey – wait a minute – look closely at the person who’s posing in each case. Why does the on-screen person look like they weigh twice as much as the person who’s in front of the camera?

Ah. I see. It’s that recurring problem again: marketing companies that don’t know where to hire skilled tech staff. Here we’ve got 3D model wrapping apparently done by someone who’s never heard of Convex Hull. This is basic Computer Science (IIRC it’s taught in almost every CS undergrad course today) – wrapping string (or cloth) tightly around a solid object is an interesting and very common problem.

OK, it’s version one; “we’ll fix that in beta”; etc. Except … you’re demoing this to:

  1. women
  2. in a fashion store
  3. in public
  4. when they’re about to pay money
  5. for clothes

…and you’re making them look:

  1. uniform weight (which for short, young Muscovites is – according to the video – usually “much larger than reality”)
  2. saggy (look at the video – they failed to register / stretch the clothing for the head/neck-to-knees length; in most cases, the women’s busts are aroudn their waists, and their waists are around their knees)
  3. masculine (most women’s clothing hangs; it’s soft, flexible; here, the Kinect models have no physics, not even primitive struct-based bending, let alone springs. Doh)

All of which put together makes this a FAIL. Technologically it’s all fixable, but from a sales/marketing perspective it’s enough to send many people screaming. Fingers crossed that the company (ARDoor) manages to make huge sales anyway – the potential here is enormous, and some of what they’ve done looks great (I like the simple interface, and the giant “Smile!” instruction).

Leave a Reply

Your email address will not be published. Required fields are marked *