Visuvi launches developer platform

Visual search engine CEO Chris Boone on taking visual search to medical and e-commerce industries

In this segment, Bambi Francisco interviewed Chris Boone, CEO of Visuvi, an image-based search engine with patent pending technology in “content-based-image recognition.”

BF: Today, you're launching a developer platform at the Java One conference to essentially broaden the adoption and functionality of Visuvi. But before we get into what you expect from the developers, talk a little bit about Visuvi. It's been under the radar. There are a couple of visual search engines that people are familiar with such as Like.com and Google Similar Search. So how is Visuvi different and how are you improving upon what they're doing?

CB: We've been in relatively in stealth mode. What we're trying to do these days is to validate to the marketplace. We do have in fact a unique set of algorithms to something very revolutionary. In visual search, what we do is referring to content based image retrieval or content based image analysis. So we're actually analyzing the content of an image and using that information to derive image results. Most technologies rely on two things such as text and color patterns. So the search companies you had mentioned - Like.com and Google Similar Search - also rely on text and color patterns. We've been revolutionary and up until this time we have been hesitant to release that to the marketplace. We are going to use the developer conference as a platform to come to the market.

BF: What do you expect from the developers now having access to your platform?

CB: As a visual search engine, we've been inundated with requests to apply our technology across the marketplace. The most notable that we have endeavored across have been medical and e-commerce. And what we are seeing now are a number of potential opportunities on the periphery and so we like to expose an API and allow folks to develop applications that leverage our core algorithms to pursue that development on their own.

BF: One of the things that you are trying to do is improve upon an iPhone application. Talk about that.

CB: Strategically, we think that on a global perspective the world is moving away from the keyboard and into more mobile type devices. Outside the United States, for example that is the predominant means of Internet activity. Look at Asia and Europe as an example. The iPhone in particular has a great interface that is simple touch, typically an index finger. So the idea there is if you can capture an image, use that image to initiate a search query, and there is no text or keyboad involved.

BF: So you're hoping a developer can improve upon this concept that you have?

CB: Absolutely.

BF: The biggest challenge for this technology is finding a business model. How do you make a business of this? What are you focused on? You're focusing a little bit on e-commerce and medical. Talk about where you think the big business opportunities are in visual search.

CB: So visual search, or in scientific lingo - computer vision - is nothing new. We're not creating computer vision. It's been around for more than 20 years. The difference is the way we are applying our algorithms to solve significant market problems. So one of the reasons why you haven't seen image-based search in medicine is because of the absence of color. So from a technology perspective, using color as a primary means of analysis is rendered useless when you're looking at the four primary means of medical imaging, such as X-rays, CT (CAT Scans), MRIs, and ultrasound, all of which are black and white. Our technology has the ability to analyze black and white images. Importantly, you're talking about five gigabyte file sizes, which takes a lot of computing resources to effectively analyze.

BF: Let's take a quick demo of what that looks like.

CB: To start, let's look at medical. As I mentioned, most current visual search technologies don't work well in the absence of colors. This is specifically something we've designed for the radiology department. In this case, we are looking at the four primary medical modalities. There's CT, X-rays, MRI, and the ultrasound. So let's say I am a medical professional and I see a white nodule here at the brain and I want to find out more information to determine if it is benign or malignant. Here are the results. We clicked on the button and submitted a search query. It came back within a matter of seconds and the results that you're seeing are other brain CTs. The coefficient that's noted here is 1.0 is an exact match. This means that it found the exact same image in the database. Moving from left to right, that coefficient goes down from 1.0 and in this case from .923. And you'll notice that this is the same patient. So, these are slices of a brain CT. But if you go down one, you'll notice a different naming convention and a different matching coefficient. This is totally a different patient.

BF: So the result here is to improve a doctor's ability to diagnose?

CB: Absolutely, in terms of two things involving medical. You want to improve patient care. We are enabling where physicians can find related patients case studies. In addition to patient care, the second thing would be reducing patient cost. We just introduced a technology that can search across billions of technologies in a very short amount of time to reduce human involvement, therefore reduce the cost.

BF: And this market opportunity is?

CB:  It's $70 billion dollar market and a tremendous opportunity.

BF: Have you signed on with Siemen's already?

CB: We are very excited to announce a deal with Bioimaging, a digital pathology company that has been funded by Siemens, yes.

BF: Let's see your e-commerce application.

CB: The idea is to capture an image, rather than trying to describe it in text, you can actually take the image from a search query. In this case, let's look at sunglasses. This is looking at one e-commerce vendor's database and you'll notice that the initial search results bring back some false positives. We show this as an advantage because the search result is initially looking for image patters, textures, and so forth. Bit if you click on something to find similar, eliminating false positives, and again by using image analysis.

BF: The idea is to allow the consumer to purchase those glasses.

CB: At the end of the day, it's about converting browsers to buyers.

BF: For you, medical and ecommerce is not the opportunity, it is general search. So talk about the opportunity of visual search.

CB: So as an early stage company, we're challenged with is how best to apply the technology. The initial market  verticals were perceived as a great way to value our technologies. We would love to be more of a Google visual search but being a visual search engine, takes an evolution, takes time, and product development. But with other products, as we continue to release, we'll make that migration toward a general purpose search. One such which is an auto tagging product which can deliver text based tags based on the content of the image.

BF: And you're going over social networks to work with them?

CB: Exactly.

BF: You're also set to close around a funding?

CB: That is in process.

BF: Good luck with the product.