Global convenience food and beverage leader PepsiCo is using cutting-edge machine vision technology from startup KoiReader Technologies, based on the NVIDIA AI platform and GPUs, to improve the efficiency and accuracy of its distribution process.

PepsiCo identified the KoiReader technology as a solution to increase the efficiency of reading warehouse labels. This AI-powered innovation helps read warehouse labels and barcodes in fast-moving environments where labels can be any size, at any angle, or even partially covered or damaged.

It operates out of PepsiCo’s distribution center in the Dallas-Fort Worth area, with plans for a wider rollout this year.

“If you find the right lever, you can significantly increase our throughput,” said Greg Bellon, senior director of digital supply chain at PepsiCo.

KoiReader’s AI-powered innovation helps read warehouse labels and barcodes in a fast-paced environment.

KoiReader technology is used to train and run deep learning algorithms that power PepsiCo’s artificial intelligence label and barcode scanning system.

Once near-perfect accuracy has been achieved, its application is extended to verify customer deliveries to ensure 100% accuracy in human-assisted picking operations.

At the Dallas plant where PepsiCo is testing the technology, Koi AutonomousOCR technology quickly and accurately scans some of the most complex warehouse labels on fast-moving conveyor belts.

It is also being researched to help warehouse workers scan pallets of soda and snacks. The same AutonomousOCR technology was also used to automate operations on the site as tractors and trailers move in and out of PepsiCo’s distribution center in Texas.

“KoiReader’s capabilities open up the potential for many use cases—starting small and demonstrating capabilities is the key to success,” says Bellon.

The system is already generating valuable information in real time, Bellon reports.

Koi’s technology can accurately track regular or irregular-shaped products, with and without labels, and calculate how long workers take to pack boxes, how many items they pack, and how long it takes them to retrieve the items for the boxes.

It acts as a real-time study of industrial engineering, answering many questions about the impact of people, processes and technology on productivity.

The broad spectrum of the NVIDIA stack is used by KoiReader in its diverse portfolio of solutions and customer workflows.

NVIDIA TAO Toolkit, DALI and Nsight Systems are used to train and optimize models on large servers with NVIDIA A6000 GPUs.

The NVIDIA DeepStream SDK, TensorRT and Triton Conclusion Server are used to maximize throughput and deliver real-time results on peripheral nodes powered by NVIDIA A5000 GPUs, and NVIDIA Jetson AGX Orin servers with module support for larger deployments.

And every aspect of Koi applications is built in the cloud using containerization, Kubernetes and microservices.

In addition, NVIDIA AI Enterprise the software suite promises to help PepsiCo confidently scale and manage its AI programs and deployments.

“The KoiVision platform was created to provide logistics, supply chain and industrial automation for enterprise customers. Our suite of solutions helps PepsiCo improve operational efficiency and accuracy in the distribution process,” said Ashutosh Prasad, Founder and CEO of KoiReader.

Subtitle: Accelerating Insights

“Today, we’re testing object and action detection capabilities as well as computer vision to figure out what kind of data we want to collect with such a program,” Bellon said.

Bellon said he’s excited for what’s next. “We’re going to travel together,” he said.