Qcify recently launched the Qcify RAY, an affordable real-time inline monitoring device designed specifically for the nuts industry.
Today, we have the pleasure of speaking with Antoon ("Toon"), one of our talented Software Engineers and the driving force behind the development of the Qcify RAY.
Join us as we delve into the journey from concept to reality, exploring the innovative features, design considerations, and technological advancements that make the Qcify RAY a game-changer in quality control and processing efficiency.
Interviewer: Hi Toon!
Toon: Hey there!
Interviewer: How are you?
Toon: I'm doing very well, actually.
Interviewer: Great to hear! We're here to dive into the Qcify RAY and get some insights into what led Qcify to create this innovative product.
What are some unique quality control challenges faced by the nuts industry, and how did they inspire the development of the Qcify RAY?
Toon: The primary inspiration for the RAY comes from the need to know the quality of products like almonds, walnuts, and pistachios at any point in the processing line. We want to be able to provide a fully integrated factory line in the future. What we were missing was an affordable, small sensor that could be placed anywhere to give real-time data on the product flow, whether it's the entire stream or a side stream. It had to be compact to fit into various parts of the line, like sizer spouts, reject lines, or accept lines—anywhere with a few centimeters of freefall. Our vision has always been to provide insights at any point in the line, and the RAY is the perfect solution for that. It's not just the nuts industry that faces these challenges; any processing line dealing with bulk products needs this. Existing solutions are often too expensive or don't provide the necessary information, making the Qcify RAY an affordable and efficient alternative.
Interviewer: Cool. So, the RAY was developed to fit into a processing line wherever there is freefall and it is also designed to be small. Were there any other key design considerations during its development that heavily impacted the final design?
Toon: Absolutely! One major consideration was the computer choice. It's an embedded device that uses very little energy, which means low heat generation and no need for active cooling. Just one coated aluminum heat sink at the bottom does the job. It operates on a switchable power supply of either 230 or 115 volts, which is really convenient. We also use a specialized camera, allowing us to keep the device compact. This camera only needs to see a few millimeters, making it much more efficient than a regular camera.
Interviewer: So, everything was designed from the start to be as compact and affordable as possible, making it easy for customers to integrate an inline monitoring system without a hefty budget impact.
Toon: Exactly! The RAY is very affordable and scalable. There's nothing quite like it on the market, especially at its price point. With its small form factor, it's a fantastic solution.
Interviewer: I can imagine this brought about some major technical challenges during development to piece everything together.
Toon: Surprisingly, the development went smoothly. We've been working on the software stack for a few years and aimed to make all our machines as similar as possible. Over time, we've built a sort of LEGO® set of components that we can assemble. For the RAY, we used a similar processing stack to the Qcify EYE. The main challenge was getting it to work on the embedded device, but once that was solved, we had a mostly working machine. Thanks to the strong foundation we already had, a small team and I were able to develop the RAY in a remarkably short timeframe. This streamlined the development and kept costs low.
Interviewer: That's great! It sounds like your previous collaborations on other Qcify solutions played a significant role in this project's success. Were there any standout points in your internal collaboration during the development of the Qcify RAY?
Toon: Definitely. Our previous collaboration on the EYE software provided a solid base. From there, the main focus was on the mechanics, which I worked on with Joachim. We built the machine around the optics, leveraging our strong core software and the prior experience on earlier projects to tackle most challenges beforehand.
Toon: Thanks!
Interviewer: So, that’s the hardware standpoint. Now, let's talk about the software, which is also based on deep learning and supports various products such as almonds, hazelnuts, and walnuts.
Toon: Exactly. For the deep learning model, we initially built separate models for the Qcify RAY, as it’s slightly different from the Qcify EYE. Over time, we collaborated with the Applications team to use the same model for both the RAY and the EYE. We calibrated the color spaces closer together and retrained the model to work on both devices. The results have been excellent, making it much easier to support new applications. Instead of building a model from scratch, we start from an EYE model and retrain it slightly. In about a week, we can have a new application ready. It’s quite efficient!
Interviewer: Wow, impressive! Is there a specific way we're leveraging advanced technology within the Qcify RAY? We've discussed the advanced hardware and software technology, but what else makes the Qcify RAY stand out in the tree nut and peanut processing industry?
Toon: The key differentiator is the combination of data processing capability, ease of integration, and cost-effectiveness. Traditional quality control systems are either bulky or expensive. The Qcify RAY processes the same data with much easier integration and at a lower price point. This makes it a standout solution in the nut industry.
Interviewer: How does the Qcify RAY improve accuracy and efficiency within the processing line?
Toon: By placing multiple RAY units along the line, you get real-time data on what's happening. For instance, bulky sorters work well until they don’t, often needing recalibration. With a RAY, you get immediate feedback if something’s off, like when the reject bin contains 95% good product. The RAY notifies you instantly to fix the sorter’s program, saving time and money by preventing the need to reprocess the entire bin.
Interviewer: Historically, nut processors sometimes only notice issues after processing an entire bin. How does the Qcify RAY help reduce waste and save costs?
Toon: Knowing exactly what you're processing allows you to price products accurately. Many customers undersell their product to guarantee quality.
In this case, you could place a RAY right before the packing stage so you can know exactly what you are selling. This ensures that, instead of underselling to guarantee quality, you can confidently sell at the correct price point. Giving the potential to increasing revenue significantly. The Qcify RAY is an income generator, as it helps you maximize the value of your product.
Interviewer: Considering its low budget impact, the Qcify RAY seems like an easy way to increase profits.
Toon: Absolutely. Even one RAY at the end of your packing lines can provide detailed data on each bin, enhancing pricing accuracy and overall profitability.
Interviewer: Does the Qcify RAY assist with industry standards and regulatory compliance?
Toon: Anywhere more data is needed to understand and manage your processing, the RAY is valuable. Whenever possible we pre-load the Qcify software with the relevant industry standard grades, for example the USDA grades for California Almonds. In any case, having precise data helps ensure compliance and improves overall quality control.
Interviewer: How adaptable is the Qcify RAY, especially within the nuts industry? You mentioned it’s easy to adapt Qcify EYE models to the RAY.
Toon: The RAY is highly adaptable. Our software backend is separate from the models, making it easy to train and update them. Placing the RAY in strategic spots, like reject lines, quickly gathers product specific defect data (mold, discoloration, shell, etc.), which is crucial for training neural networks. This data benefits not only the RAY but also the Qcify EYE and future Qcify products. It’s a seamless process that significantly enhances our ability to handle new applications.
Interviewer: So, our neural network models are essentially backwards compatible with existing data.
Toon: Yes, collecting defect data is notoriously difficult, but the RAY’s placement makes it straightforward. For instance, putting it at the reject line of a sorter could yield around 500,000 defect images in an hour. This data enriches our models and improves the overall system.
Interviewer: Customers might typically place the RAY in the processing line or at the accept stream to control product grade. However, you mentioned placing it at the reject stream to gather defect data, which is a game-changer.
Toon: Exactly. Having data at multiple points gives comprehensive insights. Traditional methods are sample-based, which is inherently flawed. The RAY, with its ability to see every object, offers a revolutionary approach to quality control.
Interviewer: Do you see the Qcify RAY influencing future trends in food processing or quality control?
Toon: Definitely. The ability to collect detailed data at multiple points enables a shift from sample-based to data-driven processing. This approach ensures precision and efficiency, potentially setting new industry standards. Also, there’s likely to be an influx of similar copycat products, indicating the RAY's significant impact.
Interviewer: We’ve talked a lot about data. Where does this data go, and how is it used?
Toon: To support one or multiple RAY units, you need a Qcify HUB, which visualizes the data and generates actionable information and alerts. Our QGrade system sends notifications if something is out of spec, providing real-time quality control.
Interviewer: Very insightful. Anything else you'd like to add about the RAY?
Toon: What really stood out was how easy it was to develop the prototype, thanks to our existing software stack. Transitioning from prototype to product took time, but the process was smoother than expected due to our robust foundation. This ease of development bodes well for future projects.
Interviewer: That makes sense. Starting from a strong base clearly paid off. Thank you very much, Toon, for sharing your insights.
Toon: You’re welcome!
In conclusion, the Qcify RAY is poised to bring exciting changes to the nut processing industry with its innovative design, affordability, and cutting-edge deep learning capabilities. The Qcify RAY offers an accessible, real-time quality control solution that can be seamlessly integrated into processing lines. Its advanced deep learning models enable quick adaptation to new applications, making processing more efficient and effective. Plus, the comprehensive data it provides allows for precise, data-driven decision-making, helping to prevent costly reprocessing and ensuring top-notch product quality. With these incredible advancements, the Qcify RAY not only enhances profitability but also sets a joyful new standard for quality control. A big thank you to Toon for sharing his insights and passion behind this groundbreaking development!
Ready to revolutionize your quality control and boost your efficiency?
Reach out to us today to learn more about the Qcify RAY and how it can transform your nut processing operations!
Contact us now to get started.
Or book a demo.
