Last week I was on a panel at the RESO conference where we talked about software personalization as an important trend. Back in 2008, I wrote some articles about improving prospecting and real estate search, one aspect of which was that we needed to get smarter about understanding consumer preferences so that consumers don’t have to page through so many listings, or can at least see the most likely matches to their interests first. As I noted in that article, the tricky part was that “there are various qualitative aspects of property selection that we don’t currently track data for at the current time.” That’s where “computer vision”, a technology that is becoming both more robust and more common, could make a difference.
“Computer vision” is the ability of a computer to analyze photos and create data out of them. Imagine a consumer likes open floor plans, modern kitchens, wide driveways, high ceilings, mature trees, or lots of natural light – those are all things a consumer might mention when describing their dream home, but little of that information is reliably tracked by agents in most MLSs. With computer vision, that data could be extracted from the listing photos by a computer as keywords by which listings could be searched – without having to manually sort through many homes and many more photos.
I recently saw an example of computer vision demonstrated by RealScout (not a Clareity client) that I was impressed by. I don’t think any company has fully leveraged the potential of computer vision and created the “perfect” product with it, but this company had clearly made some real progress on applying computer vision to real estate search. During the demonstration they showed how they could automatically tag photos so a consumer could, for instance, page through just the kitchen photos of multiple listings – even if the photos weren’t labeled “kitchen” by the MLS. The technology enables searches for normally unsearchable criteria, and to compare images for key features and rooms side-by-side and room-by-room, as illustrated below:
Images used with RealScout’s permission.
We’ve watched computer vision evolve over the past fifteen years or so – Google’s image search was launched in 2001 and has continued to get more sophisticated, and there are many other companies outside our industry that specialize in it. RealScout is not the only company using computer vision in real estate – during their demonstration they showed how Trulia has used it, and I’ve seen others explore this area over the past few years though not always release the resulting products. There are also quite a number of companies outside of our industry that license computer vision technology – of course, it would have to be optimized for real estate use. And there are some limitations to what the technology can do at this point, especially where listing photos are limited. No one wants their client to miss out on a home because the pictures didn’t highlight a certain feature. That said, this technology – if used artfully – can certainly augment existing listing search technologies and create a compelling user experience.
I have no doubts that our industry will continue to evaluate how to create a great listing search experience using computer vision, and that the number of products – both existing and new – leveraging this technology will grow over time. This is certainly an area to keep an eye on.
Share this post: