Kings River Packing group, a family-owned company with three citrus packing houses – Kings River Packing, Cobblestone Fruit, and Jireh Packing – is rolling out LUCAi™, the latest deep learning A.I. grading technology, to all of their sorting lines that currently have TOMRA’s Spectrim vision system.
King’s River Packing became the first U.S.A. pack house to roll out LUCAi™ deep learning for citrus by upgrading their existing Spectrim vision systems.
After a successful full-season trial on 20 lanes at their Cobblestone Fruit packhouse, Kings River was excited to roll out LUCAi to their other packhouses across almost 100 lanes, saying it was an easy decision.
“The performance of LUCAi at Cobblestone was outstanding. We expected to see some improvement, but the reality was far more impressive”, says Bobby Hines, CEO of Kings River Packing.
“LUCAi really does identify those hard-to-detect defects and is far more accurate overall. This will absolutely help us consistently deliver great products to our customers and provide a greater return for our business in the seasons ahead,” says Colby Campbell, General Manager of Cobblestone Fruit.
Not only has TOMRA ensured that LUCAi™ outperforms existing platforms in terms of defect identification and classification, but they also wanted to focus on enhanced usability. “The operator still plays a critical role in the process; they remain in control of the final sort parameters,” says Clinton Jeffries, Area Sales Manager, NAM.
Campbell agrees, “After working with LUCAi, our head operator was impressed. He saw that the accuracy, speed, and ease of use were a real jump forward and will allow him to focus more on overall line performance, quickly adjusting the grade when desired, and in general, trusting the consistency of the results”.
LUCAi, now available for lemons, mandarins, and oranges, has demonstrated a >99% accuracy rate for difficult defects, including rot, sunburn, clipper cut, and long stem.
“For these particular defects, LUCAi outperforms even TOMRA’s best-built traditional mapping models,” said Dr. Christopher Johnston, Head of Applications Engineering at TOMRA Food. “The issue for traditional mapping of any type – whether TOMRA’s or a competitor’s — is that it focuses solely on a few pixels that show the suspected defect. This makes it harder to achieve accuracy with classification and grading in difficult defects, like clipper cut, because it doesn’t appreciate the wider context of the fruit.”
“LUCAi is already delivering a step-change in both performance and usability, by looking at the entire piece of fruit when classifying a defect. LUCAi truly knows the difference,” said Jeffries.
LUCAi™ Citrus is the latest of a series of LUCAi™ deep learning applications.
“We’ve already seen remarkable results for Spectrim with LUCAi™ in apples and InVision2 with LUCAi™ for cherries, and now we’ve expanded to support lemon, mandarin, and orange applications,” said Dr. Christopher Johnston.
