Tyson Foods plans to expand the use of computer vision to track how much chicken moves through its plants, part of an effort to invest more in automation and artificial intelligence to cut costs and reduce waste, according to a Wall Street Journal article.
By the end of the year, the company expects to use cameras, machine-learning algorithms and edge computing, where data is processed and analyzed in near-real time without being sent to a data center, to record hundreds of thousands of pounds of packaged chicken every week.
The company has installed the systems at three facilities and plans to expand them to all seven of its “fresh tray pack” chicken plants, which churn out meat packaged for supermarket cases. Tyson has about 50 facilities in the U.S. that process chicken.
“We’re trying to apply the most cutting-edge technology in order to derive new insights and enable new ways to work,” said Lee Slezak, vice president of IT architecture, emerging technologies and analytics at the Springdale, Arkansas-based company.
The technology will allow Tyson to better control its inventory and manage the freshness of its chicken, said Lee Slezak, who reports to Scott Spradley, Tyson’s chief technology officer.
The company is investing in technology to improve operations. Over the past five years, Tyson Foods has spent more than $215 million on robotics and automation technologies, Slezak said.
Tyson, whose brands include Jimmy Dean, Hillshire Farm and others, said that profit rose to $557 million in its latest quarter, up 1% from a year earlier, as growing U.S. chicken production put a lid on prices. Tyson’s beef sales slipped on 2% on weaker volumes, but sales of chicken increase 6%.
The inspiration for the computer-vision technology at Tyson came over a year ago after Lee Slezak and his team visited an Amazon.com Inc. cashierless checkout store in Seattle. Lee Slezak and his team of data scientists and technologists sought insight from machine-learning experts at Amazon’s cloud services division and launched a pilot program last year based on similar technology.
The system identifies the type of product, such as a package of chicken thighs, and the stock-keeping unit, or SKU number for the batch of products, using computer vision. An automated scale records the weight of a batch of chicken packages in the cart. An operator looks at a nearby screen and confirms the weight and SKU number.
The accuracy rate for identifying the product type and SKU number is in the high-90% range, using computer vision, an estimated 20% improvement over manual processes, Slezak said. Humans still validate some of the computer’s inventory tracking and train the algorithms, Slezak said.
The computer-vision systems could also help detect foreign objects at facilities that process products at a high volume, which will be useful for food safety, Slezak said. “We’re trying to apply technology everywhere we can as a company to drive down costs and drive up efficiency and business value,” Slezak said.
