fbpx

Chasidy Rae Sisk

chasidy

Chasidy Rae Sisk is a freelance writer from New Castle, DE, who writes on a variety of topics.


She can be reached at crsisk@gmail.com.

 

 

Monday, 16 November 2020 18:05

A.I. Photo Estimate Algorithms Advancing, Still No Replacement for Shop Blueprinting

Written by
When the A.I. detects the possibility of underlying damage, “we recommend that the owner take the vehicle to the shop for a more in-depth assessment,” Tractable North American Automotive Operations Head Jimmy Spears said. When the A.I. detects the possibility of underlying damage, “we recommend that the owner take the vehicle to the shop for a more in-depth assessment,” Tractable North American Automotive Operations Head Jimmy Spears said. Courtesy of Tractable

Index

Insurance companies have been pushing photo estimates for more than five years, and collision repair shops have been objecting to the inaccuracies of these estimates for the same amount of time, because these virtual claims typically require a supplement---or five!---due to the potential for inaccuracies inherent in the process.

Artificial Intelligence technology attempts to combat these challenges by a more precise estimate using artificial intelligence.

 

Tech companies use algorithms to enhance the accuracy of photo estimates, which reduces the cycle time, making A.I. photo estimates appealing to both insurers and consumers. After being trained in image classification, the algorithms are able to identify damage, send claims to parts-pricing software and produce an estimate in seconds.

 

“These algorithms are very different to how people used to do computer vision, because you actually get the algorithm to figure out the right patterns in the object,” explained Alex Dalyac, founder and CEO of Tractable. “Instead of telling the A.I., ‘This is what a front bumper looks like; look for a corner like this and pixels like that,’ you feed the algorithm millions of images.

 

"Some contain a front bumper and some don’t. On a rainy day, a dark day or a sunny one; an undamaged bumper; or one that needs three hours of repair. And the algorithm itself figures out the best combinations of pixel patterns that give it the most accuracy. It’s kind of magical, but it’s very data hungry.”

 

Algorithms are also able to learn quicker than human estimators, and Tractable has trained their A.I. by inputting approximately 10 million photos of damaged vehicles for the algorithm to analyze. Of course, vehicles have become increasingly complex due to technological advances, so simple repairs aren’t actually simple anymore---how many sensors need to be inspected before a repair technician can replace a bumper?

 

Although the A.I. can be trained to identify that there may be damage to components that are not visible in the photo, it cannot provide a complete estimate for these repairs; disassembly and blueprinting...


Previous Page Continue reading »

Read 451 times