Why real estate valuation needs Computer Vision?
Ashley Do | 3min read
We are living in a world where technological advances are rapidly developing and the application of artificial intelligence is no longer an option. Instead, it is essential to increase the competitiveness of businesses in any field including real estate. Soon, real estate valuation models that do not apply AI to a certain extent will soon become obsolete. According to Rob May, Talla’s CEO and Co-Founder, “It is surprising that some companies are slow to accept AI because they are not sure where to apply it or whether it will work well. I think they don't understand that until they find out, they will be left behind”.
Automated pricing models (AVM) are often used by financial institutions to make decisions about everything from equity lending to credit card limits. It is not just a business of houses, but understanding the value of each house is very important to minimize losses and manage credit risk. AVM applied worldwide should be a lot better because AI technology can improve 15% of price prediction as a sign that traditional models have been underperforming in recent years.
Why is the home valuation based on traditional models no longer good enough? They are based on old data, digital data, census data, IRS data and state/regional sales data that may not fully reflect the investment into the house. Even combining all of these data sources can sometimes produce an incomplete picture. Although traditional AVM models are very fast and cheap, they cannot consider the quality and status of an asset without looking at the asset, thus limiting the accuracy of a key tool in financial services industry.
Besides, traditional computer algorithms prefer to process using formulas instead of images of houses. None of the fields on the list is called "a missing shingle on the roof". To get this right, the AI model must look at the image and understand how that image will affect the price. More specifically, the model needs to learn about the room it is looking at, understand the texture, colors and indoor and outdoor objects. Think of this solution as an innovation on a dating site where no one has ever been able to see people's profile pictures, instead they can only see their height, weight, and other biomedical data. Clearly, a picture is worth a thousand words.
The problem of integrating image data into home pricing is really appropriate for the AI solution. The images on the list can tell us the story of property valuation that can be enhanced by traditional numerical, categorical, and text data associated with a real estate listing. For example, the granite face shows a certain investment of money for the house. Subtle elements like the condition of the paint on the wall also tell us a lot about the expected selling price of that house.
Research by Foxy AI - an AI application company in the field of real estate has now launched a new AVM method that combines computer vision and deep learning to assess the quality and condition of real estate. The entire system is displayed through APIs - Application Programming Interface and products use advanced training techniques and in-house predictive modeling sets to classify rooms, so that the image set helps determine the quality as well as the condition of the subject assets, thereby improving the accuracy of valuation. This will probably be the future solution that brings about high accuracy and efficiency for financial institutions.