Honey.AI – The Evolved and Optimized AI-related IoT solution for the honey industry
Idea and relation to VEDLIoT
Pollen analysis in honey industry is used to determine the floral source and authenticity of this highly valued food. It’s a very manual test which involves studying a honey sample in a glass slide with a microscope for 1-2 hours, so all the existent pollen grains species are identified and counted by a widely trained laboratory technician. Manual counting is expensive, time-consuming, involves human error, and the results are obtained deferred. However, these analyses are essential and mandatory in the honey industry, since are used for quality assessment, fraud detection, and labelling legal directives. Honey.AI is a novel equipment that allows on-site real-time cost-effective IoT device to honey industry stakeholders to assess their honey’s quality with robotics, computer vision and deep learning tools.
Honey.AI will contribute to the VEDLIoT ecosystem extension with mainly 4 aspects:
- Offering a specific use-case for the agrifood sector.
- Shifting to a state-of-the-art edge-hybrid distributed architecture to drastically reduce the GPU cloud computing currently used, as well as assure lower latency, higher reliability in remote locations, reduce traffic, bandwidth, and data storage.
- Providing a high demanding AI/ML application, with significant computing needs, and with 2 years of previous data gathered with cloud computing at AWS.
- Promoting the project and the VEDLIoT technologies in a diverse number of events from different sector.
Objectives
- Understand VEDLIoT technology and get familiar with the modular HW, the toolchain and accelerators. Obtain useful and reliable support from VEDLIoT experts.
- Adapt Honey.AI’s current system to the new architecture, including recheck of the digital camera and touch screen/display compatibility, reducing the sending/capturing frequency of the embedded system, redefinition of control board.
- Re-code the App to make it compatible with the new acquisition hardware and to compile in the new ARM architecture, as well as change the GUI to make it usable without a full OS running in the background (functions to configure, WIFI, network, OTA updates, power control etc.),
- Adapt the neural network to be deployed in the embedded hardware to get the maximum performance per watt using VEDLIoT technology by means of using Kenning and EmbeDL.
- Migrate the pollen detector from current AWS cloud computing site to the specific HW selected, after detailed benchmarking done of the different alternatives available. Also integrate the crystals, yeasts, color, and starch detection functionalities.
- Design, manufacture and assemble the new version of Honey.AI, test it in real operational field with early adopters and compare its performance with the current version.
- Negotiate terms and conditions for future exploitation with VEDLIoT partners in case of enhanced performance of Honey.AI 2.0 and advantageous cost-benefit analysis.
Approach
Cheap imports of counterfeit honey from other regions (specifically Asian countries, not meeting the quality standards) are seriously endangering European beekeeping, and the consequences for food production are severe. These cheap imports drive the prize of honey down making it very difficult for beekeepers to make ends meet, which has serious consequences for beekeepers, local ecosystems, and consumers. The number of beekeepers is starting to decline in Europe, since the producers cannot compete with Asian imported honey prices, and the consequences are very serious, since bee population surveillance totally depends on the health of beekeeping sector, an activity which is of vital importance to the European economy, and environmental sustainability. In Europe, the Commission has recently created a labelling directive for honey to include the countries where the honey has been imported from, after 16 member states called for a revision.
Introducing Honey.AI solution for automated honey quality monitoring has the potential to allow European beekeepers to add high value at the beginning of the value chain, thus differentiate from low-priced foreign honey imported from Eastern hemisphere, meeting current market demands. An increase in economic profitability for beekeepers would help to revitalize the domestic industry and have a positive direct impact on well-being and stability of apiculture, where low returns usually cause beekeepers to abandon the sector. With Honey.AI it is estimated to reduce potential Honey fraud (mislabeling of honey type) by 20% and increase profitability of beekeepers by 1- 4%. In consequence, our proposed solution will be able to contribute to apiculture sustainability improvement in a realistic manner.
Honey.AI presents an on-site solution for automated honey quality monitoring, providing a disruptive technology that allows the honey packer or beekeeper to self-analyze their honey samples at their premises, at real-time, without the need to pack and send a sample of each honey drum harvested or purchased to a specific external laboratory. Considering an average of 400 honey samples analyzed per honey packer (or even 1000 samples for honey cooperatives) and bearing in mind that there are only few specialized laboratories in EU who carry out pollen analysis, the avoidance of thousands of unnecessary logistic trips between diverse long-distance locations assures significant CO2 emissions savings for the environment.
Expected Impact
Honey industry is an activity of vital importance to the sustainability of Europe, and plays an important role in agriculture, since apiculture and bees are responsible for more than €30 billion a year in crops, being responsible for more than 30% of world’s food by providing pollination in about 1000 flowering species, therefore impacting on the whole food supply chain.
SUSTAINABILITY IMPACT
- Helping to prevent honey fraud, while improving sustainability of beekeeping sector: With Honey.AI it is estimated to reduce potential Honey fraud (mislabeling of honey type) by 20%, and increase profitability of beekeepers by 1- 4%. In consequence, our proposed solution will be able to contribute to apiculture sustainability improvement in a realistic manner.
- Environmentally friendly: Honey.AI presents an on-site solution for automated honey quality monitoring, providing a disruptive technology that allows the honey packer or beekeeper to self-analyze their honey samples at their premises, at real-time, without the need to pack and send a sample of each honey drum harvested or purchased to an specific external laboratory. Considering an average of 400 honey samples analyzed per honey packer (or even 1000 samples for honey cooperatives), and bearing in mind that there are only few specialized laboratories in EU who carry out pollen analysis, the avoidance of thousands of unnecessary logistic trips between diverse long-distance locations assure significant CO2 emissions savings for the environment.
In terms of creation of employment, it is foreseen that Honey.AI will generate over 15 job positions over the first 5 years of operation.