12 Mar There is more to a photo than you see
Images can tell you more about a property than what you see and this technology can help buyer search.
Topic – The opportunities we overlook
Kevin: This morning on the show, Kylie Davis is going to catch up with a gentlemen she met at Inman Connect in New York and an interesting talk about some of things that we miss in photographs. Here’s Kylie.
Kylie: I’m here with Dominik Pogorzelski from Restb.ai which is a really exciting company here at Inman New York City doing work around AI and extracting data from photographs and visuals. So welcome Dominik how are you? I hope I got your surname right, I’m sorry.
Dominik: It was good.
Kylie: I got close. So tell us about how much, or what sort of data is sitting inside photos? What can we, what more can we get out of a photo once we’ve got one.
Dominik: I think it’s a bigger question and that’s one of the things that shocks me the most, is that the real estate industry is sitting on millions of images.
Dominik: And as far they know, for them it’s just jpeg one, jpeg, two and jpeg three.
Dominik: They have no idea what’s in these images, what are these images, and there’s a treasure trove of information in those images.
Dominik: As you mentioned. Images can tell you things like the condition of the house, what specific features are in the house. You know, is it a, is there a kitchen island, is there a granite countertop, or is it you know, just a regular countertop. Is there a fireplace, a pool. So all these very enticing features about a house that users look for when shopping for a house.
Kylie: Yeah. So I guess what we’re seeing is this move away from unstructured data, which is sitting inside a photo. And when we were talking before you called it like a black box, that photos are like black boxes.
Kylie: And I’m a bit of a data geek, so this idea that you have something that before has been really flat and just a thing and suddenly we can start to pull data out of it and then use those features, like kitchen benchtops or European appliances or…and then start to use them as the search criteria because that becomes part of the data feed, right?
Dominik: Exactly and I love the way you put it. It’s a way of structuring unstructured data.
Dominik: Extracting from the images and then surfacing out to the user…
Dominik: Or the broker in order to then be able to control those images. And basically regain control of their own images.
Dominik: In order to help the home buying experience.
Dominik: And like you said, imagine a scenario where you’re searching for a home and you put in the criteria ‘I’m looking for a place in this neighbourhood for this price and this square footage’, and then you get the results of all your listings. And then you can take that a whole step further and add a whole visual layer to it and the user can say ‘Well I’d like to see these results by the kitchen photo. Or I’d like to filter by kitchen with a kitchen island that has a granite countertop’, for example. And then you can really start engaging the users in a much more visual way, which is important because the home buying experience is a very emotional experience, right? And what better way to engage with humans on the emotional side than with visuals…
Kylie: With something visual.
Kylie: Because you really are gonna start to search for something by what it looks like as opposed to what it sounds like in the description, aren’t you?
Kylie: That’s really really cool. So, and I can’t imagine ever, that photos have so much data in them, that I can’t ever imagine a real estate agent on the planet who would sit down and type into like a spreadsheet or into a data collection format all of the different bits and pieces in it. So how, tell us about Restb.ai. How are you guys doing it?
Dominik: So we built a proprietary artificial intelligence technology that basically scans the images and then does a bunch of different things. So it can then firstly, it can classify the image into the type of room that it is, so it’ll say this image is a kitchen. And then it can extract pertinent and relevant information from that image. So it can say that this image has a kitchen island, stainless steel appliances, you know, dark cabinets and light hardwood floor, for example. So it can really go in and understand the content and the context of an image. So then basically as you said, structure that unstructured data.
Kylie: But it can also start to identify what kind of condition it’s in too, can’t it? As in whether it’s in excellent condition, or poor condition, or things like that, which adds like another layer over the top. Because suddenly you can understand from the data, or you’re suddenly able to pull a data stream around what things might need to cost to repair or things like renovators are getting a lot more stuff around that.
Dominik: And then that can also add, as you said a different layer. Well I’m looking for a fixer upper. I’d love to buy a house, fix it and flip it.
Dominik: Well then that, you know, has to do with the condition.
Kylie: Or I’m looking for a house I just want to fix the kitchen. I’m okay fixing kitchens.
Kylie: But I don’t want to do bathrooms, or I don’t want to have to, you know, add a back room or things like that. So you’re using AI to do a lot of this stuff.
Dominik: Mm-hmm (affirmative).
Kylie: And you guys are still fairly new aren’t you? Been around how long now?
Dominik: So the company was founded three and a half years ago.
Kylie: Yeah. So what I think is really exciting about this stuff you’re doing is that you’re making available to agents, like they’re sitting on this really rich source of data, and in real estate it’s all about who owns the data.
Dominik: Mm-hmm (affirmative).
Kylie: But the only people who own the data are the people who can extract the data. So how do you guys work with brokerages or with franchise groups?
Dominik: So it’s pretty simple. So a broker or a franchise group or a CRM would basically just send us the image, we, that image gets processed through our AI, and then we send back the meta data from a GSM format, back to the source. So we don’t sell the data, we don’t own the images…
Dominik: All that remains on the broker side. We just send back the information of here’s what we picked up that we think is in this image.
Kylie: Wow. And so how long does it take to do all of that?
Dominik: It’s less than a second per image.
Kylie: Wow. Really?
Dominik: And you can do multiple images at a time. So you can go as high as a hundred to four hundred images per second if you want.
Kylie: Wow. And so who are you working with over here, or what kind of business are using you?
Dominik: So we’re working with MLS’s, with data providers such as Tribus, for example.
Dominik: One of the main data providers here in the U.S. We’re working with FBS, the maker of Flex MLS, with photography companies like VHD…
Dominik: So it’s really a broad spectrum of people who have some sort of touch point with images.
Kevin: And who want to have that data to then use it for their own purposes.
Kylie: Yeah. Fantastic. Alright, well look Dominik, thank you so much for your time. I think that’s really fascinating that we’re now, that we’re…It’s like we’ve sort of struck a seam of gold, isn’t it?
Kylie: That it’s been under our noses, all this time…
Dominik: Mm-hmm (affirmative).
Kylie: And we just haven’t had the tools or the resources to extract the value out of it. But now, suddenly photography which is like an inherent part of every real estate agent’s toolbox is gonna start to be an amazing source of information that really enhances sort of the experience for buyers and sellers going forward. And I guess also becomes an asset that agents are sitting on that they can now get more value out of.
Dominik: Hopefully. That’s what we’re striving for.
Kylie: Yeah. Great. Fantastic. Dominik, thanks so much for your time.
Dominik: My pleasure. Thank you very much for having me.
Kevin: Tomorrow on the show I’m going to feature a talk that I had with Tammy Bennnell. Delightful person from Exit Real Estate. That’s an interesting model. That’ll be coming up tomorrow morning. See you then.