Skip to main content

News

Oceanit uses AI Chip Innovation to Win Hawaii’s First Annual ‘AGathon’

Posted January 26, 2018 in

On January 13th, the ‘Harvest Vision’ team took first place in Hawaii’s first ever agriculture-tech hackathon: AGathon.

Oceanit teamed up with Kauai Coffee Company and Kamehameha Schools to develop an innovation that can ‘see’ the ripeness of coffee cherries utilizing an artificial intelligence microchip. By determining ripeness, Harvest Vision could improve Kauai Coffee’s harvest values by more than a quarter million dollars per harvest.

“Kauai Coffee Company is the largest coffee producer in the United States, and coffee is Hawaii’s second most valuable agro crop, last year making about $62 million dollars,” says Fred Cowell, General Manager of Kauai Coffee. “The sad stat is that each year, (we) lose about 20% of profits due to harvesting inefficiency.”

Coffee growers must harvest quickly and efficiently because coffee cherries are ripe for a very limited period before becoming overripe. In response, growers typically use mechanical harvesters as a way to speed up harvests, controlling the speed and shaking intensity of the machinery to dislodge coffee cherries.

While mechanical harvesting speeds up the process, it also limits farmers’ ability to control the quality of their harvest. The machines, by themselves, cannot distinguish the good cherries, so monitoring the progress of the harvest requires well-trained inspectors to subjectively characterize the condition of harvested coffee.

Before harvesting, an inspector will determine if the field is ready for harvest based on collection from a single, representative row. Once harvesting starts on a field, the quality of the harvest can only be inspected between ten hour harvesting shifts, when collection trucks are brought to the processing plant. This creates a large lag between when farmers discover they may be harvesting an unripe (or overripe) field and when they can react.

When there is a high quantity of unripe or overripe coffee cherries, these cherries must be discarded resulting in a lower harvest yield and increased waste.  If farmers could get real-time feedback on coffee ripeness as they harvest, they could reposition harvesting machines, speed up or slow the harvest, or even adjust the harvester settings, shaking less unripe cherries from the branches, maximizing the harvest.

Artificial Intelligence and ‘Harvest Vision’

The innovation behind Harvest Vision’s sensor technology is an artificial intelligence (AI) chip that can used to optimize the yield of coffee harvests, by “training” the AI to recognize ripe cherries vs. unripe cherries and informing farmers when too many unripe cherries are being harvested.

Harvest Vision shortens the feedback loop, because the AI system can evaluate in real-time to provide consistent and repeatable data. Farmers are able to prioritize harvest areas through computerized data analysis, and because of that, they are able to obtain higher yields of high quality coffee.

The image above shows the team’s demonstration during the AGathon competition, using the AI technology to determine the ratio and location of “ripe” vs “unripe” jelly beans – simulating coffee cherries. When the system sees too many unripe cherries on the plant, it can alert the farmer to adjust the mechanical harvester to shake less, leaving more of the unripe fruit on the plant.

By the end of the competition, Kauai Coffee Company had determined to work with Oceanit to develop Harvest Vision further and intends to test the new system during the September 2018 harvest.

AI Chips in Agriculture

Harvest Vision could prove to be a game-changing shift in the future of agriculture. Beyond using the AI chip to inspect coffee harvests, it could be adapted for the processing and inspection of a variety of other crops.

 “The greatest thing [about the technology] is that you can train it. The AI-based system can learn to recognize different shapes, features, colors, or textures depending on the crop”, says Sumil Thapa, Materials Engineer at Oceanit.

Popular fruits such as strawberries and blueberries are among the crops to have potential for significant increased yields via accurate inspections. Better yields decrease production costs and waste, ultimately making farming more economical and sustainable.

Artificial Intellgience could be built into platforms like drones to allow remote inspections that go beyond simple, visual analysis of crops.  It could be integrated with devices such as sprinkler, irrigation, or harvesting machines to adjust settings in real-time, saving energy and water resources.  AI could also be used to optimize manpower, determining where and when farmers should be going to certain fields to maximize efficiency and yields.

The value in AI chips is just starting to be realized in the United States. As AI develops, it is becoming apparent that new kinds of hardware are needed to optimize AI’s performance. The implementation of Harvest Vision’s specific AI chip is unique, in that the chip doesn’t require cloud computing, making it possible to perform practical advancements in processes on a hardware level.

“The AI chip is small and independent, it doesn’t rely on the cloud or the web. It learns what ripe coffee looks like, it provides feedback for each operator in real time, and links them together so you can see when one is underperforming,” says Fred Cowell.

Oceanit hopes to use artificial intelligence chips far beyond agriculture, becoming a multi-application tool that can be used to augment many of Oceanit’s breakthrough technologies. As Oceanit opens new frontiers in smart sensors, the IoT, machine learning & cognitive computing, health & fitness, energy and beyond, the hope is that low-cost AI hardware will unlock many new avenues of exploration.

“Something that we were originally looking at with the AI chip was to utilize it for face recognition,” says Ian Kitajima, Commercialization Director at Oceanit. “The technology could allow easier access through homes and automobiles, with higher protection and privacy.”

On January 24, 2018, Sumil Thapa from the Oceanit AGathon team visited Bytemarks to discuss Harvest Vision, to listen to the podcast visit: http://www.bytemarkscafe.org/2018/01/24/episode-491-ag-hackathon-winners...

To learn more, please contact Oceanit.