Consistency has always been the benchmark of successful cherry grading. While modern vision grading systems have delivered remarkable advances in throughput, sizing accuracy and defect detection, maintaining consistent results across changing fruit conditions remains one of the industry’s ongoing challenges. Artificial intelligence is now helping address some of those challenges, not by replacing conventional vision grading, but by making it even smarter.
For many years, vision grading has relied on sophisticated camera technology and colour-based image analysis to assess every piece of fruit moving through the packhouse. These systems have transformed the industry, allowing packhouse managers to achieve levels of speed, consistency and objectivity that simply aren’t possible with manual grading alone.
However, cherries are naturally variable. Fruit colour, stem colour, surface finish and defect appearance can all differ between varieties, orchards and growing regions. Even within a single production run, these characteristics can change enough to influence how a vision grading system interprets an image. While today’s systems perform exceptionally well, achieving the best possible results has always relied on understanding and accommodating this natural variation.
One of the most important elements in that process is accurate stem detection.
The stem might appear to be a small part of the fruit, but it plays a significant role in how the rest of the image is interpreted. If a grading system cannot reliably distinguish the stem from the fruit itself, colour-based analysis becomes more susceptible to false detections or inconsistencies when assessing defects. Accurate stem identification provides a cleaner, more reliable starting point for every grading decision that follows.
Rather than relying solely on colour spectrum analysis, GP Graders’ radiai™ platform uses artificial intelligence to recognise stems based on learned visual characteristics. This enables the system to identify stems more consistently despite natural variations in colour and appearance, creating a stronger foundation for downstream defect analysis while maintaining the sizing accuracy growers and packhouses have come to expect from modern vision grading.
Building on this foundation, radiai™ incorporates a range of AI models, each trained to recognise specific characteristics of cherry quality.
Rather than attempting to classify every imperfection using a single approach, individual AI models focus on specific defect types including open wounds, healed cracks, bruising, browning, wind rub and ring splits. Additional models identify waste fruit, doubles, multiple fruit and residual flowers that may remain attached after harvest. By allowing each model to concentrate on recognising a particular characteristic, the system is better equipped to distinguish between defects that may appear visually similar but require different grading outcomes.
The result is not simply more defect detection, but more consistent defect detection. By reducing ambiguity in image interpretation, AI helps deliver grading decisions that remain reliable across changing fruit conditions, supporting greater confidence in packhouse quality standards.
Importantly, these AI models are not intended to replace proven vision grading techniques. Instead, they complement existing technology by addressing grading challenges that have traditionally been more difficult to resolve using conventional image analysis alone. AI becomes another layer of intelligence within the grading process, working alongside the established vision grading platform rather than replacing it.
For packhouse managers, the benefits extend beyond the technology itself. More consistent grading decisions mean greater confidence that customer specifications are being met, fewer unnecessary adjustments during production and more repeatable results from the first bin through to the last. As labour pressures continue to increase and quality expectations become more demanding, consistency has never been more valuable.
Artificial intelligence continues to attract significant attention across agriculture, but its longterm success will ultimately be measured by practical outcomes rather than technical capability.
In commercial cherry grading, that means helping packhouses make better, more consistent decisions every day.
At GP Graders, AI is proving its value by enhancing an already proven vision grading platform in gpVision™, giving packhouse managers greater confidence that every cherry is being assessed as accurately and consistently as possible. As the technology continues to evolve, its greatest contribution may not be replacing traditional grading methods, but quietly improving the decisions that matter most.
