Door Anne Bruinsma
Results WUR Life Sciences Hack
The very first WUR Life Sciences hackathon was an initiative from several bachelor and master studies. The goal was to have creative tech talent work on targeted challenges offered and mentored by representatives of Ag Tech companies. The hackathon was supported by the European wide Smart Agri Hubs program. With a big thank you to these companies that carried the event with us, we present you the results of the first WUR Life Sciences hackathon!
On the menu for the event were four different tracks. Teams had 32 hours to understand the challenge, deepdive data, think of a data driven solution and then run it, build it or prototype it. There were 4 challengers, and 6 + 1 teams. You can find their presentations and more detail in the WUR Life Sciences Hack channel on the FarmHack Forum, including photos and video.
First Prize Winners
The winning team, team Betsie, was one of three that worked on the Lely & Rovecom challenge. The dataset they provided combined nutrition data with cow health data from milking robots on several thousands of individual cows.
Building on a correlation matrix they analysed the data to show relevant relationships, such as the predicted daily milk production, fat/protein ratio and mastitis. These insights could already be of value if offered to the farmer through early warning signals. They identified roughage and water intake as possibly key data to take the feed/milk analysis to the next level. They elaborated on facial recognition possibilities to gather such type of data. The jury granted them € 5.000,-.
Some results of the predictive modelling done by team Betsie
Second Prize Winners
Runner Ups were team Unisearch, working on the Unilever challenge that any researcher faces: the ambition to find key information in a context of ever expanding amounts of scientific literature, too little time and limited search functionality offered by current search engines. Their solution was to include full text in their search engine, and reinforce learning (A.I.) by building an interactive search engine. The creativity, technical skills and their zealous commitment impressed everyone that had the opportunity to see them work. They were awarded with €3.000,-. Some even said they hacked Google Scholar, and topped IBM Watson. We might see them grow into a start-up with funding and guidance from Unilever Venture Funds!
Prize winners of the first WUR Life Sciences Hack
Third Prize Winners
The third prize was awarded to another team working on the Lely/Rovecom challenge, called DairyBite. They put in a lot of work to compare different statistical models and tools for data modelling. They scored how well they performed on the data provided by the challengers. They won hack bonus points for doing a real live demo in their presentation, which showed a dashboard to monitor cows and visualize several indicators related to both milking and feed data. They were awarded with €2.000,-.
Everything was measured, even the teams own performance 🙂
Honorary Mention: BeVine
A special prize (the famous Farmhack Hoodie) was awarded to our Bulgarian friends from BeVine, a student team that participated in an earlier SAH hackathon and won participation in the WUR hackathon. Their goal is to prevent diseases in vine production, by developing a disease prediction model based on leave-image recognition in the vineyard. During the WUR hackathon they worked on integrating meteo data and fine-tuning their business plan.
It is always very hard to not being able to award all participating teams at a hackathon. Often the differences between them are very slim. So with just as much regard, here is an overview of the remaining teams:
Also working on the Lely/Rovecom challenge was the all student team LeCom.
They looked into the possibility to forecast health information based on feed ration data. They did statistical analysis to search for correlations with a series of indicators for sick cows. They found correlations with VEM, Crude Fibre and OEB. The team was firmly rooted in ag practice, bringing in some highly valuable ‘Boerenverstand’ into all this hocus pocus!
Working on the NIVA challenge, around the use of machine data by government, was a team consisting of representatives of the Dutch paying agency (PA) on the one hand, and a mix of API experts, scientists, software developers and a student on the other. Although they did not manage to build an actual data pipeline, feeding machine data into the current webservices of the PA, the team did (for instance) identify key issues for the PA to deal with when consuming machine data and develop a processing chain to use machine data for identifying plot boundaries (to replace the manual process of drawing plot boundaries). Very relevant results all around (check their final pitch).
Machine data: translating binary data to point geometries and the section data to polygons.
Working on the Hendrix Genetics challenge was the team HenGen Chicks, deepdiving a unique and unusually rich dataset on both phenotypes and the different environment in which such phenotypes have to perform. Their challenge was to take this complex interaction into account and predict animal performance. They quickly ran into data issues, needing to think of quick ways around missing data, unclear data or. It is important to stress that also ending up with clean data, better understood data and insights on data can be a very valuable end-result of a hackathon!
Partners and sponsor of the 2019 WUR life Sciences Hackathon