Case StudyKewpie

Kewpie, a Japanese food manufacturing company, used deep machine vision that identified defective potato cubes on the production line with the same level as accuracy as humans

Kewpie, a Japanese food manufacturing company, trained Convolutional Neural Networks (CNN) to identify defective versus quality potatoes. A video feed from the production line was monitored and the network could identify defective potatoes with a similar accuracy to humans. This resulted in savings of more than $100,000 per production line. The project took six months.

Context

Kewpie, "the food manufacturing company often considered to be the number one in food quality and safety in Japan." They wanted to use machine learning to detect defective potato cubes on their production line.

The Project

"Its system monitors the video feed from the production line, and makes a sound when it detects a defect." They used Google''s TensorFlow with convolutional neural networks to identify potatoes. The project took six months.

AI Usage

Convolutional neural networks implemented on the TensorFlow ML framework.

Data

Training data sets were likely to be labelled images of defective versus quality potatoes.

Results

They "achieved similar level of accuracy as human inspectors... Kewpie saved more than $100,000 per production line in removing the need for inspection equipment. All of this only took them 6 months before seeing results. "

Related Use Cases

Use Case

Predict potential quality issues with products through visual recognition

→

Use Case

Detect defects and quality issues during production using visual and other data

→

←Back to Case Studies