Retail owners using artificial intelligence to detect shoplifting in real time
PORT HURON, Mich. (CBS DETROIT) — Forbes recently surveyed retail store owners and found that nearly four out of five lose between $500 and $2,500 a month due to shoplifting.
Some of them are turning to artificial intelligence to catch shoppers in the act.
Shoppers who feel they can sneak a buy-one-get-one-free deal past the counter might have a little more trouble doing so past a specific set of cameras. The cameras have AI recognition technology installed, alerting the owner if someone tries to grab a snack without paying for it.
"Veesion is artificial intelligence that you plug into existing cameras to detect gestures associated with potential shoplifting. So, people putting items in the jacket, trousers, backpack or purse for example," said Veesion co-founder Benoit Koenig.
Veesion has entered 4,000 stores worldwide — 500 in the U.S. — and that number is growing.
Its algorithm analyzes several different cameras simultaneously, looking for subtle cues in a way the human eye cannot keep up with.
"The first component is human detection with the red or green square you can see around each person," Koenig said. "The second component is called pause estimation to estimate the position of all the body parts. The third component is called object recognition to detect a backpack or purse instead of a shopping bag or shopping cart. All those components together will provide you with the probability of any gesture at any time."
Since the technology is solely analyzing body movements, Koenig says Veesion is better at avoiding bias when detecting shoplifters than those behind the counter.
"The algorithm doesn't care about what people look like, it cares about gestures. It looks at body movement and will pick up on a gesture to place an item in the jacket, in the trousers, in a backpack, or purse," he said.
If someone tried to pocket some chips without paying, Koenig said, "That will generate a small video alert on the mobile phone of the store owner in real-time so they can actually ask the client if they need help, if they need a basket so people know that they're being watched. It probably deters the attempt before it actually happens."