Snap Fashion – coming to a small screen near you!
Jenny here. Wow, it’s been a busy two weeks for Snap Fashion. The highlight has to be filming for BBC Click. We’re really excited about it being aired, and I keep joking that our editor Sarah is about to embark on a great modelling career, having had her body shape discovered by the app on camera.
Kate Russell, who got in touch with Snap and filmed us on the day, was asking some great questions about Computer Vision which got me thinking – how much do the public know about Computer Vision, and how much do they actually want to know?! So, here’s my “Computer Vision in a nutshell” guide.
Computer vision is a way of making a computer “see” in a way that we expect humans and animals do. Sounds simple, doesn’t it? The trouble is, all that we’re given in a photo and no other information. Once again, that doesn’t sound too hard. The only problem is we don’t appreciate how complex our brains actually are. Seeing things is all about context. We know that that circle is a football because of its markings, coupled with the fact that it’s being chased around by a load of grown men. A computer, however, sees groups of white and black pixels bunched together on a vaguely green background with lots of coloured splodges. Not so simple now.
Therefore, in my opinion, the essence of Computer Vision is context. How do we make a computer see and comprehend? This is the crux of why computer vision hasn’t been out there for years, and why everyone is still struggling to find the optimal solution. Here at Snap Fashion we’ve developed algorithms which will focus on Computer Vision for clothing. The context of searching for clothing on a red carpet is very different to searching for logos, buildings or album artwork, which is why Snap Fashion is launching in such a specific domain, and why other products, such as Google Goggles, are focusing on different areas of Computer Vision.
But is it a bad thing that these solutions are so specific and segmented? Not in my opinion. As long as it results in the best results and user experience for consumers, why should it matter? We’ve just got to be mindful of how we integrate the technologies in the future.
Finally, I’ll leave you with this thought. It is very hard to impress consumers with a computer vision product – one bad result and your product is instantly denounced as “not working”, “not quite there” or, quite frankly, a bit rubbish. Now, think about one of the things that you first learnt to do. Before you could even talk you were pointing and recognising. Babies point at the toys that they want to get hold of. Cats can recognise their cat bed (or a lovely pile of newly washed clothes) to go and nest in. Pigeons can tell the difference between cigarette butts and tasty crisp morsels. The functionality of Computer Vision products isn’t seen as intelligent but just as common sense leading, rightly so, to highly unforgiving testers!
But of course Snap Fashion is amazing… you’ll just have to wait a little while longer to see the results for yourself!
Have a good week, and enjoy the rest of our blog.