Fruit Sizing App

Spectre sizes 9 million apples during US harvest

By December 2, 2020 No Comments
Reading Time: 5 minutes

Hectre’s computer vision fruit sizing tool, Spectre, was in full use during the recent US apple harvest. With the ability to capture and sample approximately 125 apples per bin with accuracy levels of 95%+, US apple warehouses and orchards were busy clicking away as the harvest rolled on.

slice image of Spectre size distribution chart

Hectre introduced Spectre commercially to the US market for the first time this year and uptake was hugely positive with eight fruit warehouses using Spectre during the US apple harvest

The computer vision tool, which provides users with accurate fruit sizing in seconds from the simple click of an iPhone or iPad, detects all visible apples from the top layer of the bin and then estimates full size distribution, serving up a user friendly distribution graph which includes critical data such as average count size.

Included in the eight US fruit warehouses were leading businesses, Washington Fruit and Produce Company, and Yakima Fruit and Cold Storage. Both businesses, which operate out of Washington State, have quality and innovation as two of their key foundations for market advantage.

Caption: A behind the scenes snapshot of Hectre’s computer vision fruit sizing tool, Spectre, as it begins detecting apples from the top of the bin during the fruit sizing process, before the distribution chart is served up to the user.

Image showing apple detection by Spectre

“It was extremely rewarding to see how the US market utilised Spectre during this year’s apple harvest, and even more impressive was how they took the improved data they were capturing and applied it to their storage and packing management decisions. This is exactly the value we foresaw when Spectre was at ideation stage, so it’s fantastic to see that playing out. We’re really excited to fully bring Spectre to the NZ market now and give our Kiwi customers the same benefits experienced in the US,” says Hectre CEO, Matty Blomfield.

Washington State is the leading producer of apples in the US with a whopping 65% of apple production share and 95% of apple export share. Growers had to battle extreme windstorms, wildfires and a deep freeze during the 2020 apple harvest, as well as the impacts of a global pandemic, so this year’s harvest was an overly stressful time for all involved. Despite the challenges though, more than 60,000 apple bin images were captured by US Spectre users resulting in more than 9 million apples sampled for size. 

Photo of Luke Butters R&D Manager at Hectre

Luke Butters who heads up Hectre’s R&D team stated, “It felt as though all the effort over the past 18 months had paid off, where growers and warehouses were getting real value from the tool. Reducing decision making stress during a manic harvest, reducing human error rates due to subjective sampling and lack of sample size, and arming packhouse sales teams with greater confidence from reliable data, are all benefits we were hoping to serve up to our customers with the introduction of Spectre.” 

Valuable learnings have been taken from the US harvest experience and are being used as the base for additional Spectre enhancements as the Hectre team continue their agile approach. A key iteration which has already been developed further and is now available for Spectre users in NZ and worldwide, is the ability to take bin images and group them into lots to match truck receiving, providing greater time efficiencies and lot size distribution data. Mounted automated image capture is also being trialed in the US to take advantage of their open roof trucking approach where top bins are visible and video capture is also underway. 

Sampling for colour will also be available for the upcoming NZ harvest and then work on defect detection will follow. And good news for those growing and managing other fruit types, with work streams progressing at pace for Spectre to sample size for both citrus and pears.

Hectre’s Spectre journey began as an idea back in 2018. After the Hectre team surveyed NZ growers for market need and interest, many months of image capture and detection development followed in order to train Spectre to recognise apples accurately. Harnessing computer vision and machine learning technology, plus buckets of passion, the creation of the Spectre prototype came next. 

A flight across the Pacific Ocean and a mighty road trip saw the team meet with Washington State growers and fruit warehouses as the Spectre prototype was demoed and additional data captured. Supported by a group of NZ growers who shared Hectre’s passion for quality and innovation, Hectre piloted Spectre during the NZ 2019 apple harvest, gaining valuable feedback and enabling further improvements before the US commercial introduction.

If you’d like to book a Spectre demo, or discuss how Spectre might benefit your business, we’d love to hear from you. 

Bitnami