Back to Industry Trends & Insights

Computer Vision: Seeing (and Understanding) is Believing

Cv See Beliving

Computer vision will be instrumental in automating document processing…here’s what you need to know

By Peter Piela, Ph.D., Head of Solution Development

Wikipedia defines Computer Vision (CV) as “an interdisciplinary scientific field that deals with how computers can be made to gain high-level understanding from digital images or videos.”

And getting computers to understand what they are seeing is a challenging task. In her 2015 TED Talk “How We Teach Computers to Understand Pictures” Fei Fei Li, a Computer Science Professor at Stanford University who is also Co-Director of Stanford’s Human-Centered AI Institute and the Stanford Vision and Learning Lab, explains that humans don’t see with their eyes, but with their brains, and that there is a big difference between “seeing” and “understanding” (what you are looking at).

As humans, we benefit from thousands of years of evolution and the knowledge that we have gained throughout our lives to know and understand what we are seeing. In recent years, even our most advanced technology struggles with the challenge of understanding images or videos. As Li said in her TED Talk “Even the smartest machines are blind.”

However, advancements in CV are helping us make great strides in the automation of document recognition and processing, and this will make a big impact in the mortgage industry.

Why Mortgage Document Processing Needs Computer Vision

The mortgage application process involves many different documents and systems coming together to produce data. And the verification and validation of all that data takes a lot of time.

It’s a labor-intensive process that involves gathering, processing and storing information and data, and all of this must happen before any analysis can take place. You need a lot of labor to process documents, and there is often a lot of delay throughout the process as you wait for the next step to be completed before you can move on to the next one.

The hope for the future is to automate much of the mundane parts of this process to make it faster and more accurate, freeing up humans to focus on things humans are uniquely good at: helping customers and building relationships that can grow the business.

The Impact of Computer Vision = Better Automation

CV plays a key role in making this advanced automation a reality. Using CV means that software does not have to read each and every character in a document. It can simply “look at it” and recognize what kind of document it is, then zoom in, extract and use the relevant data, or make note of when certain data is missing or incorrect. This is a big improvement from older technology, such as Optical Character Recognition (OCR), which extracts every character that it sees on a document. CV is faster and more efficient.

It’s important to know that making CV a reality requires a lot of up-front, behind-the-scenes work. Building a CV system for the mortgage industry requires large amounts of data in order to train the system about the documents it will process. And it can be a very wide range of documents – sometimes more than 300 different ones per loan. And loan documents can vary from state to state and city to city, creating even greater diversity and complexity.

Banks don’t have access to all of those different documents – they can only work with the ones they generate themselves, and they don’t have the resources to build a CV system with a deep range of knowledge. This is where a third-party company like AI Foundry comes in. We are building an enormous training data lake by aggregating and anonymizing data from many different lenders. We get a lot of varied data to work with, and as a result the model grows quickly. In the end, everyone gets to benefit from the vast knowledge that the system has gained.

To get to this point with CV, it’s taken us almost two years of hard work to gather data and build the system, curate hundreds of thousands of documents and teach the system to recognize them. But the end result is that lenders will have a much more automated document processing system, which will allow them to process a higher volume of mortgage applications, do it faster than ever before, and reduce the cost of doing it. CV will have a significant impact on the mortgage industry, with lenders and borrowers both seeing benefits.

For more information on AI Foundry’s Computer Vision, check out our new eBook entitled:  
Unleashing the Power of Computer Vision on Lending: Higher Volumes. Lower Costs. Loyal Customers.
Download ebook+