Guest: Munjal Shah - Riya

Munjal Shah - Riya
Hi! I am Munjal Shah, I am the CEO, Co-Founder of Riya. I am going to show you today our next product called Like.com, and it's a new Visual Search Engine where for the first time you can search with a photo instead of searching with text. So, let me show you this. So, this is a new website Like.com. We allow you to really search for products and specifically products that are static in nature, so jewelry, handbag, shoes, watches and in another week or two we'll also let you search for clothing.

But let me show you this, you can start a search in number of different ways, one is you can click on one of our celebrity pictures here, a lot of people do this with magazines like, People Magazine etcetera, and what it does, it then runs the similarity and shows you all the boots that if thought looks similar to the one that Tyra Banks is wearing. It's not perfect, but actually in this case, the imperfections if you will I'll give a little variety, so you can actually see some differences.

Let's do another search here. So, we can also just start by doing a text search. So, I'm going to search for a Chronograph Watch, it then finds all these watches that are tagged chronographing, here its not doing a similarity search yet or a likeness search. Let's say, we really like this one, so now let's do a likeness search and say, finally things that look similar. You notice it loads real fast, which was one of the key technology things we did, and it found all these watches that it thought really looks similar. So, steel and black, it kind of realize that's the colors of this and the general shape, but still give you a little bit of variety. As you go down the screen, it lessens the sense of similarity, so it starts showing a few things that are blue etcetera.

Now, one of the key features of like, if the ability feed zero on certain features that you really like, so, if you like this watch and you really like the fact it has three dials, you can draw this box around the section and say, "You know what like, find me just things that have three dials in it." It says, "Hey! Well, may I get the color, the shape of the thing you clicked on." I am going to say, 'shape' in this case. It reruns a search, re-looks at the visual signature and if you notice it now shows jewel watches, even this one has in it three dials, if you look very carefully, has three dials there, these have three dials, these have three dials, this one does. So, what it did is, it really went through and found this detailed feature and what you do it. You can do this for buckles, you can do this for certain patterns, you can do this a number of different factors.

Let me show you another shoe search here. So, can now search for Red Strappy Shoes, and again it just does a text search, just a kind of get you oriented, to let you find to what you're looking for. Now, let's say we like this pair here and we do a likeness search. Now, here the actual shoes it thought looked similar. Now, strappy shoes are actually quite a difficult thing for a computer algorithm because just a white space in between it's very hard to define, where things start and end, and what it got things that are pretty looking similar in terms of being strappy, there's a couple of things like this one and this one that are not quite the right thing but still very similar. Now, if you want the shoe though in a different color, let's say teal or a bluish kind of green, you click this, it reruns the search for you, it still focuses on shape and pattern but then changes up the color and shows you all these shoes that look similar but now in a different color.

So, those were two examples, I'll show you the third example. You can even search for, lets say, Summer Shoulder Bag, and if you do this search, there's this bag here that's kind of ugly, but I am going to pick it because it has a lot of pattern to it, and so, this algorithm went through looked at hundreds of thousands of bags came back and said, "These are the ones that looked the most similar," and if you think about it, it kind of got it, which is that it realizes all these little, little pieces and so it is this one, and so it this one, and so it is this one, and found all these bags that have little, little pieces and even though the different colors, but lets say, you really just wanted the color and the shape, the color and the pattern of this to match, you didn't care so much about the shape. You can use our sliders, dial up color, dial down shape, dial up pattern, it resorts all your results and pulls everything up that have the same color and pattern but different shapes now. So, that's the real power of the system is that -- it's not just, "Hey! Do a likeness search and we give you, what we give you, you can really fine tune it, you can really focus it, you can even fine tune it and focus it by all the normal things you expect from a text site where you can see by price range, by what brands are there, by what sites it came from. I only want stuff from Zappos or Amazon, we you have 2,000 user results came from Amazon, 6,000 from Zappos, by Style, I want shoulder bags or totes etcetera. So, you can focus all of that.

I'll show you last thing in example here in jewelry. So, in jewelry its -- lets search for Emerald, and it shows you things that it thought were emeralds and lets say, you actually really like this ring. You can do a likeness search on it. Now, here just looking at the whole shape, so its not only looking at emeralds, although we might have a way for you to tell it, "I only want emerald to the final version that we put out," but it shows you all those different rings that kind of have similar settings, but lets say, you just love this green color of this emerald. Again you can do a detail search focus on the box. The box on the stone say, "Search by color," and now it says, "Ignore the rest of the setting," just show you all the different emeralds in this color, and it goes through and finds them all. So, just lots of ability as was fine tune and detail. Now, of course, if you want you can just start with these celebrity pictures, click on them, find lots of things that look similar as long as they're not too obscured it'll work, in some cases the celebrity's hand is across it and it won't work, some cases it will.

So, those are some of the examples, so you can even do one more here, I misspelt it but, we'll still find it. So, here is a -- this kind of unique stone looking bracelet, it runs a likeness search on it, and it comes back here with things that it thought look similar. The first three match well, we actually probably don't have too many of these, so starts degrading after that, but you can then click on this one and say, "Wow! Let's take a look at this one" And I found a lot more and you can say, "Well, no, I really like this one," and it just keeps iterating and showing you more and more that look real similar. So, again, I mean how would you capture in text that this look like this, you would never, I mean you just wouldn't capture that. This cold/blue effect combined with the fact it's kind of glassy, and it's a glass sort of shaped item and we're able to find that because of the really using appearance. We've done some neat things here with the breadcrumbs visual because you don't remember the names of the items you click on, so this way you can go back, and over time will even let you upload a picture that you take as well as get our browser toolbar to be able to start a search from anywhere on the web, and that's about it. A lot of this about browsing, so we let you turn it up if you want and show 50 or 100 at the time and if you have a big screen and you have 23 or 30 screen it will give you the whole thing in one day also.

So, that's Like.com, I hope you guys like it and use it. It's just the beginning for us. We really believe in the ideas searching by photograph and we're going to continue to build other categories, home and garden categories, clothing as well as landscape pictures and even other things on the web. So, this is really we as first installment towards Visual Search focused at products as a starting point. Thanks!