David A. TeichSenior Contributor
B2B technology analyst, marketer, and consultant
Much of the focus on deep learning systems for vision has been in three areas: autonomous vehicles, facial recognition, and robotics. However, as with the many other areas of artificial intelligence (AI), vision will have a far wider impact on society than in those three areas. The logistics of relocation are heavily depending, no surprise, on what is being moved. Vision can be applied to that challenge in order to create more accurate estimates much faster than before.
Moving is stressful. As a one news article points out, “about one in five Americans (23 percent) think that moving is more stressful than planning a wedding, according to new research. Twenty-seven percent think it’s more stressful than a job interview, and more than one in 10 (13 percent) even go as far as to say it’s more stressful than a week in jail.” Given that census data shows more than ten percent of Americans sixteen and over move every year, that’s a lot of stress. However, moving isn’t only stressful for the people relocating.
As someone who has moved a lot, including overseas, I understand the challenges experienced aren’t just those of the customer. Think about the costs of providing an estimate for moving a residence – as that is simpler than the large picture of corporate moves. Most moving companies are small and local, yet they need to find a person to go out to every household, to walk through, write down the property, then create an estimate that drives a contract. To make things more complex, moving is seasonal. For instance, Summer is a very busy times for many reasons, including families try to wait until children finish school years. A company will find it hard to staff experienced estimators for a short period.
The concept of using the nearly ubiquitous smart phone’s camera to help customers quickly take an inventory from which to create an estimate is attractive. Taking pictures, though, isn’t sufficient if it means an estimator has to squint at an image and identify everything.
One company using AI to address the estimation challenge is Yembo. The company decided to tackle this problem using deep learning. It was clear that supervised learning, where the trainers know the answer and feedback information to the system to make it more accurate. One amusing example of early learning problems was teaching the system to differentiate between white doors and refrigerators. Another part of training had to do with objects near other objects. If, for instance, the system isn’t sure about a chair, but it’s near a kitchen table, percentages for an initial classification can be increased.
Data, as with all software, matters. Yembo used a mixed dataset, both from open sources and custom imagery, to help identify many objects, including things partially obscured by others. That allows a customer to take an image of a bedroom and recognize a nightstand partially blocked by the bed. Another fun example was based on the fact that many people have TV’s near fireplaces. Early training had the system then identify fireplaces as TV’s. Negative training had to be performed to show the system that those weren’t televisions.
When the team began to trial, another great advantage of applying AI to consumer solutions rather than engineering solutions came up: a natural focus on the user experience (UX). If the customer is going to walk around a home with a smart phone, the application needs to be simple. A single button to press is easier than selecting menus. In addition, downloading an app can be an annoying extra step for a process that happens only once every few years. The designers made the system to run in a browser so there’s nothing to install.
“Yembo’s software lets our customers conduct a virtual inventory in just a few minutes,” explained David Cox, executive vice president, JK Moving Services. “This innovative AI tool is enabling quicker turn-around; a more intuitive, accurate estimation process; and an ability for our estimators to handle more business while still increasing customer satisfaction.”
Another added benefit exists, as shown by the testing. Images are much more useful than a list to the physical movers, the driver and crew who have to show up to pack items and load a vehicle. Boxes, padding, and equipment can be more efficiently allocated to each crew, and the accuracy of time estimates is also improved alongside volume and weight estimates.
Vision AI has a number of widely ranging applications. A key focus of this column is to notice how AI is moving from very academic and technical sandboxes into the world of business. What Yembo is doing interests me as it is a clear example of looking at a real problem and addressing it by leveraging AI as a component of a robust system.