
Artificial Intelligence
Discover how TOMRA Food develops technologies from across the spectrum of artificial intelligence to enhance sorting precision, improve food safety, and maximize yield. Our intelligent systems learn, adapt, and make smarter decisions — empowering food producers to deliver consistent quality with greater efficiency.
Optical sorting machines engineered for artificial intelligence
The power of any artificial intelligence model comes from the data used to train it.
TOMRA's high-spec hardware combined with our global footprint across commodities and geographies means your business benefits from models trained on high quality data.
We’re not only training our algorithms with the best possible data, but we’re generating clear data for your algorithms to read. And with stellar data comes new levels of performance for your operations.

Types of artificial intelligence used in TOMRA food sorting and grading machines
Machine learning
What is machine learning?
Machine learning in post-harvest food sorting and grading is the use of AI algorithms to automatically analyze and classify fruits, vegetables, and other food products based on factors like size, color, ripeness, structure, defects, and contamination.
How does it work?
Machine learning aims to learn patterns without being explicitly programmed. It’s a statistical modeling technique that uses algorithms to automatically learn from the data that’s generated by your crops running through machine.
Depending on your machine and your application, your algorithm will work one of three ways:
Unsupervised learning: Your machine learns a pattern to split characteristics into groups.
Reinforcement learning: Your machine learns through trial and error to maximize a reward.
Supervised learning: Your machine features the most sophisticated type of machine learning and uses labeled data to learn a pattern and assign to given labels.


Deep learning
What is deep learning?
Deep learning models are deep, complex, global models that begin to correctly classify hard-to-read defects from the moment you switch them on in your packhouse.
Our primary deep learning product is LUCAi™, and its superpower is looking at the whole piece of fruit to predict the most correct classification.
Where machine learning can accurately classify a product by assessing a few pixels, LUCAi™ considers the entire piece of fruit. It can not only assess the following but assign the correct grade based on your tolerance level:
Quality: We can classify produce based on quality references, such as Premium, Standard, or Reject.
Ripeness & maturity: We can determine ripeness stage by color, texture, or infrared analysis for Brix values.
LUCAi™ can see whether an organ is underripe, rip, or overripe.
Defect & disease:
LUCAi™ easily spots and assesses bruising and mechanical damage, pests, scarring, and mold and fungal infections.
Size & shape:
LUCAi™ can identify roundness, size, and deformities to classify each piece of fruit.
Dive deeper into LUCAi™

LUCAi™
Right product, right pack, season after season: AI at TOMRA
Setting the right grading tolerance takes discussion, time, effort, and labor – and the one thing that remains the same is that the next season will be different.
TOMRA Food’s AI capabilities extend across our entire active product line. From deep learning to simple machine learning applications, our AI features and capabilities support your strategic decision-making and lift the burden from your operators.
The result: right product, right pack, season after season – no matter where the year takes your business.
Where will we go next?
TOMRA Food aims to extend its deep learning algorithms across the commodities we serve – and go further.
We’re constantly working towards offering more precision to our customers with improved usability, so operators can raise the bar on best performance.
One thing we can predict is constant change in the global food industry, and TOMRA will be here every step of the way to support our customers, existing and new, through this journey.


