By teaching computers how to interpret the visual world, MLS photos are starting to unlock an enormous treasure trove of listing data that was previously unavailable.
While the number of bedrooms, bathrooms, and square feet may have historically been the primary way to search for real estate, this may soon just be your starting point. What if your criteria could include the size of the backyard, quantity of natural light or the vastness of the view? Alas, most of that information is not included in current MLS data systems. However, recent advances in “computer vision” now make it possible to analyze listing photos and automatically identify hundreds of features, providing thousands of additional data points to you and your clients.
Offering consumers the ability to search at a new and personally relevant level was the impetus for RealScout to pioneer the use of computer vision within the real estate market. This new “machine learning” capability is unlocking unique new functionality in the home search process – functionality no pre-2016 homebuyer has ever had.
What is Computer Vision?
Computer vision is the science of teaching computers to learn and understand the visual world. Humans can easily glance at photographs and accurately identify and describe what we see, such as room type, layout, lighting, appliances, etc. It is, however, only through recent advancements in technology that computers have developed similar skills. It’s the same technology that Google’s self-driving cars use to distinguish pedestrians from bike riders and countless other details in a dynamic road environment, and which Social Media platforms use to suggest that the person in the photo next to you is one of your friends.
Real estate listing photos carry vast amounts of valuable information for homebuyers, all of which was previously unsearchable. But now, RealScout is unlocking information such as consumers’ neighborhood-specific features and trends based on true interests and preferences discovered from their search behavior. When paired with an agent’s expertise, this technology is setting a new standard in the consumer homebuying experience. And those REALTORS® using the best technology and machine intelligence will be able to successfully guide their clients through the increasingly complex homebuying process.
Real Estate Search Explodes
Basic online real estate search is simply the tip of the iceberg of an innovative user experience. Now with computer vision unlocking a treasure trove of visual information, vast amounts of property data have become searchable and consumable.
Think what would happen if photos could provide the information necessary to answer the following questions:
- What is the exact size of every room?
- What is the architectural style of the house?
- When will I need to repair/replace the roof?
- Which direction is the family room facing?
- What are the fine-grained feature classifications: granite/marble/concrete/tile countertops?
- What plants are in the front yard?
- How much snow will I have to shovel in my driveway?
Answers to questions like these will open up a new world of refined search and decision-making capabilities for REALTORS® and their clients.
In 2012, RealScout began manually training their computer models with mind-boggling amounts of data in order to build a vast database tailored to real estate. They used human taggers to label more than 7 million listing photos with dozens of elements ranging from room type to features, attributes, and other key information; a challenging task.
Teaching computers how to understand pictures is significantly harder than it might seem. In real estate, it’s currently accepted that almost 100,000 manually tagged photographs are needed to help a computer identify the nine different room types in a standard house: kitchen, dining room, living room, bedroom, bathroom, etc. And that’s only recognizing the room type. Thereafter, each subcategory of data—floor types, wall finishes, kitchen cupboards, etc.—require tens of thousands of additional hand-annotated photos. The various scenarios in real estate are in the tens of millions and the variations in the many billions.
In 2015, RealScout took this enhanced property information and applied a unique computer vision functionality called “Compare” to enable side-by-side displays of photos of particular rooms or of different listings, enabling buyers to compare properties on multiple dimensions.
In the future, RealScout will take computer vision to new levels by enhancing other elements of the real estate sales process. For example, property insights extracted from photos can power highly detailed CMAs for pricing properties or enable high-accuracy look-alike search for agents searching for homes with similar features. Even information like square footage, construction quality, and neighborhood conditions can be extracted through computer vision techniques, increasing the transparency of information for agents and clients.
What Does This Mean For REALTORS®?
Computer vision is still in its early evolutionary stage, and the full impact is yet unknown. However, it is already proving to be an invaluable tool for empowering REALTORS® to enhance their value proposition. Providing a better understanding of data and, more important, facilitating informed decisions are subtle and highly personal skills that machines cannot acquire. REALTORS® who effectively integrate these technologies will be better able to meet and satisfy their clients’ needs, and will naturally be in greater demand.
In the not-too-distant future, computer vision might even be used to automatically predict days on market and sale price, or suggest to listing agents how they can optimize the sale price. One thing is for sure: The industry again finds itself on the cusp of a rare opportunity to take the lead in adopting innovative technologies to improve the consumer homebuying experience.
The Internet started to redefine the home search process and now, just 20 years later, computer vision is redefining the process once again. The future of real estate search has been forever changed.