At car auction houses, AI is a tool, not a worry

The process uses a gantry, an overhead structure with high-resolution cameras, which works like a drive-through, Huang said. Vehicles will be photographed as they’re driven through the gantry, and the process gives a user two sets of output — the vehicle images and, in the future, an automated damage-detection layer on top of those.

“Once the gantry is built, it takes the images and then the artificial intelligence can be layered on top,” Huang said.

Human inspectors will still double-check to verify vehicle conditions, Huang said. There are some images that can’t quite be captured — such as hail damage, Huang said.

“But for the large damages, the most common damages, it’s actually quite accurate already,” she noted.

The technology is advantageous because it may help root out damage “false positives,” according to Manheim. Because the pictures are taken from multiple angles while the vehicle is moving through the gantry — producing a 3D image of the vehicle — such phenomena as reflections can be better recognized, Manheim said. With 2D pictures, a reflection might be misinterpreted as damage, the company said.

Manheim will deploy more gantries over the next 18 months to two years, depending on how quickly supply chain issues affecting cameras and microchips level out, Huang said. Some Manheim auctions will get one gantry, some will get multiple gantries and some won’t get any.

But Fyusion’s imaging software is at all Manheim locations already via handheld devices. Auction locations that don’t get an AI-enabled fixed gantry will still use that 360-degree imaging from Fyusion.

Manheim did not say exactly how much it’s paying for Fyusion or each gantry, citing ongoing costs, but said it invested $300 million in the last three years to improve vehicle information and the Manheim Marketplace.