LUCAi™ Severity Score: the little number that gives your operator full control of your sort 

Your line, your customers, your business. How do you push the limits of your packout and your profit margins?

Optical sorting systems gave you the first step: automation and accuracy. And artificial intelligence took those features up a notch to allow you to run faster and smarter. But LUCAi™’s deep learning model offers something the others don’t: serious granularity.

LUCAi provides your operator and business very fine differentiation for certain defects and blemishes – and this is its superpower. The model doesn’t achieve this through detection and classification alone: fine differentiation gives your operator complete control in the form of the Severity Score.

What is the LUCAi Severity Score?

The LUCAi Severity Score is a rating of how bad the model predicts a defect or blemish to be. When combined with your tolerances, it keeps a piece of fruit that belongs in first grade out of your second grade packs.

Let’s take a step back and remember how LUCAi processes an image. Upon scanning from Spectrim, the deep learning model will assign each piece of fruit a:

Scarring: First or second grade? LUCAi knows the difference thanks to Severity

Citrus scarring is a prime example of where a Severity Score can make real improvements to your packout – and your profit.

LUCAi for Citrus arrives in your facility already knowing the difference between rot and scarring. It gives you a confidence score to that effect. But scarring may not matter depending on the severity of the scarring or the pack. So LUCAi assigns it a figure between 0 and 100 – with 0 being a minor blemish and 100 likely sending that piece of citrus to Juice.

We built the Severity Score because we know that blemishes and defects aren’t always what they seem: scarring becomes a problem when it’s too severe to be kept in the pack. But the bigger, more difficult problem that Severity Scores solve is when the scarring is so light that it belongs in first grade – but traditional mapping or hand grading sends it to second grade. 

So, when your head operator sets your tolerances, they’ve getting not only the accuracy of detection. They’re maximizing your packout. It’s all part of getting the right product in the right pack – and not giving away a single piece of citrus.

​