Work
opacity arrow
Dr. Oetker Computer Vision
Case Study

How Machine Learning can help make the perfect Margherita

Using hi-res cameras, computer vision and AI, we’ve helped global frozen food giants Dr. Oetker identify overtopped, uneven or substandard pizzas, without a single human being in sight.
Award winning work
No items found.
Challenge
How could technology and Machine Learning help Dr Oetker make the perfect pre-made pizza?
Solution
We added four high-resolution cameras to the Dr. Oetker production line, capturing images of each pizza as it gets being baked, sauced, topped, packaged and frozen. The images are processed by Helm Engine’s Machine Learning service and then sent to the User Interface (UI), where the operational team can easily track data on overtopped, undertopped and uneven pizzas, as well as production line efficiency.

Putting Helm Vision™ to work

Using a massive amount of visual data, we used Helm Vision™ to analyse the production line in real time and identify patterns that are similar to all pizzas in order to create a 'model' pizza.

Putting Helm Vision™ to work

Using a massive amount of visual data, we used Helm Vision™ to analyse the production line in real time and identify patterns that are similar to all pizzas in order to create a 'model' pizza.

Eliminating objectivity. Increasing productivity. Reducing waste.

By utlising the 'model' pizza, our solution is able to accurately detect whether a pizza is up to standard, and thereby taking human error and objectivity out of the equation, putting these decisions, as well as tasks like labelling, in the hands of Machine Learning.

large image 1large image 1

"

We have always been committed to surprising and delighting our customers with pizzas that taste and look like a pizza prepared from scratch. However, monitoring for uneven or overtopped pizzas is open to human error, and quite a challenge, as one person’s under-topped pizza is another’s idea of perfection. We are working with Helm to help the decision-making with machine learning – with great success so far."

Tobias Bauer


Senior Executive Manager of Dr. Oetker’s Pizza Production Unit

Keep learning

Helm Engine’s Machine Learning service groups all the pizzas with similar properties together. It’s able to tell which pizzas are overtopped and using too much cheese, or which pizzas are uneven and compromising the quality of the product. The machine learns more with every pizza that passes, and will only get better over time.

Keep learning

Helm Engine’s Machine Learning service groups all the pizzas with similar properties together. It’s able to tell which pizzas are overtopped and using too much cheese, or which pizzas are uneven and compromising the quality of the product. The machine learns more with every pizza that passes, and will only get better over time.

The User Interface

The User Interface (UI) is compatible with multiple devices. In this case, the data is displayed on a TV on Dr. Oetker’s production line, which gives them the data they need to track pizza production and quality.

large image 2large image 2

Result

With over a terabyte of data to work with, we’re currently perfecting Dr Oetker’s cheese station before rolling it out to more stations, more lines and more countries in the next few months.

1 TB

Data captured by cameras

50 K

Pizzas monitored per day

Give Helm Chat™ a go in real time

Powered by Helm Engine™