The cherry grading, sorting, and packing industry is undergoing a radical transformation, driven by artificial intelligence (AI). From optimizing sorting systems to enhancing quality control and even predictive maintenance, AI is making packing lines faster, more accurate, and more cost-effective. This evolution is particularly impactful in high-speed, high-volume cherry processing operations, where precision and efficiency are paramount.
Challenges and Goals
Cherries are a highly variable product with quality, colors, shape, size, and of course the
long stem, all presenting an array of options that compound together to create a complex object to analyze. Being a somewhat delicate, small fruit makes cherries more susceptible to damage. Common defects include cracks, rot, insect damage, bruising, wind scar, sunburn, and stem decay, as well as issues like softness, pitting, and stemless cherries. Like many fruits, cherries have a semi-permeable membrane (cuticle) that allows water to enter and as such, extra care and precaution need to be taken to prevent cracking.
The packhouse goal, therefore becomes... beyond the orchards, how do we maximize the returns on the good fruit we harvest?
The good news is that further to all the progress that's been made with vision grading systems
over the last two decades, a new solution has been presented as the industry embraces Artificial
Intelligence.
Imagined, trained, and ready to outperform
The term, 'Artificial Intelligence' was coined in the mid-1950s and has actually had a couple of
false starts for widespread uptake. However, AI's recent surge in popularity is driven by deep
learning, a biologically inspired neural network that leverages vast datasets and the immense
computational power of modern technology.
"That's where GP Graders' new radiai journey began. Recognizing the vast amount of information we already had from our years and years of cherry grading and letting the latest technology go to work for us again," said Stuart Payne, Director at GP Graders.
"Over the past five years, GP Graders has meticulously indexed and annotated hundreds of thousands of images of cherry defects, building one of the most comprehensive datasets in the industry. This extensive database forms the foundation of the radiai technology, enabling artificial intelligence to detect and classify defects with remarkable accuracy. Unlike traditional rule-based sorting systems that rely on predefined parameters, radiai utilizes advanced convolutional neural networks (CNNs) to automatically recognize and differentiate between a vast range of defects without explicit programming. By continuously refining its understanding through machine learning, the system adapts to the natural variability found in cherries, improving both precision and efficiency.
"At the core of this technology is AI's ability to associate subtle visual patterns with specific defect categories, mimicking the expertise of a trained human grader - but at an unparalleled speed and scale. For instance, unusual discoloration may indicate underlying bruising, while uneven surface textures can signal cracks or scarring damage. Irregular shapes such as cherries with spurs or double growths are instantly flagged for reclassification, ensuring only premium-quality fruit makes it through the grading process.
"By leveraging deep learning, GP Graders' radiai continuously refines its defect detection
capabilities, making your cherry grading more accurate, consistent, and adaptable to the evolving challenges of high-speed packing operations."
The Proven Impact of AI
Operators using gpVision with radiai are now reporting near-perfect stem detection, regardless of how many stems appear in a single image. This breakthrough eliminates a common challenge in cherry grading - misidentifying stems from neighboring cherries, which can distort size measurements and show defects where they didn't exist.
By accurately isolating each cherry and its respective stem, radiai ensures precise diameter calculations, leading to superior sizing accuracy. This also translates to a higher cup fill rate, as cherries can be processed more efficiently without sacrificing grading accuracy. Unlike traditional systems that can struggle to maintain accuracy when scanning large volumes, radiai seamlessly handles increased throughput, enhancing both productivity and packing efficiency.
"A major improvement in defect detection comes from radiai's ability to distinguish stems from actual fruit imperfections. Previously, the wooden or brown portion of a stem could be misclassified as a crack, leading to the unnecessary rejection of perfectly good cherries.
With AI-powered refinement, this issue has been significantly reduced, minimizing false
positives and preventing unnecessary waste.
"In the past, each misclassified cherry meant a direct financial loss, as high-quality fruit
was mistakenly discarded as a false-positive reading. By improving classification accuracy, radiai ensures that only genuinely defective fruit is removed from the batch, maximizing pack-out rates and revenue for growers."
One of the most critical aspects of cherry grading is crack detection, as even minor surface
damage can affect a cherry's classification. However, not all cracks impact quality equally
- location matters. radiai now has the intelligence to differentiate between various crack
placements, allowing for more nuanced grading. A small crack near the bottom of the cherry,
for example, may still allow it to be classified as seconds, while a side crack, which is more likely to affect the fruit's integrity, results in rejection. Additionally, the system evaluates stem quality, identifying dehydration or browning, which are key indicators of freshness and shelf life.
"One of the most groundbreaking advancements using gpVision with radiai is its ability to
detect fresh splits, a notoriously difficult defect to identify. These micro-tears in the cherry's
skin are often too subtle to register on traditional grading systems, as the freshness of the
wound lacks sufficient contrast in standard imaging. However, with access to an expansive and continuously evolving dataset, radiai has nearly eliminated this issue. By recognizing even the most delicate defects, the technology gives growers greater confidence that only pristine, high-quality cherries make it into cartons and punnets, ensuring customer satisfaction and reducing costly returns."
Reducing machine operator load
With manual sorting now practically a thing of the past, the opportunity to improve the machine operators' experience and reduce the time it takes for new operators to learn and optimize the system presents itself.
With radiai technology now at the core of gpVision, the operator experience has been completely transformed. What was once a complex, data-heavy process requiring meticulous number tweaking has evolved into an intuitive, user-friendly interface. Instead of manually adjusting intricate settings in complex spreadsheet formats, operators now have access to streamlined controls, including easy-to-use sliders and visually intuitive configurations that simplify the entire grading process.
"Compared to other systems on the market, gpVision is already recognized for its superior ease of use. Now, achieving exceptional grading accuracy is easier than ever—operators simply review the data, make minor adjustments using the intuitive sliders, and let the AI-powered system handle the rest. This seamless interaction not only reduces the learning curve for new users but also enhances efficiency, allowing packhouses to optimize sorting precision with minimal manual input. The result? Faster setup, consistently high-quality grading, and a significant boost in productivity."
Conclusion
While the world comes to terms with how AI will affect our daily lives, the trailblazers are exploiting this monumental shift in technology to seriously enhance workflow, production, quality, and ultimately profits.
With defects manifesting in countless ways across every cherry, only AI has the capability to
model and recognize these variations in real-time with unparalleled accuracy.
Having processed more than 10,000 tons of fresh cherries this past season alone, packers
across the southern hemisphere, running gpVision with radiai technology, conclude the results were spectacular. The grading outcomes were more accurate than in the past, the identification of defects otherwise unable to be seen was groundbreaking and the ease of use was celebrated by operators.
gpVision with radiai is the future of cherry grading and we invite you to explore the difference it could make in your business.
For more information:
Stuart Payne
GP Graders
Tel: +61 3 9585 9444
Email: info@gpgraders.com
Website: www.gpgraders.com