The latest SFN hackathon proved that developing workable, innovative solutions to some of our grandest challenges is entirely possible within a short timeframe – you just need creativity, cooperation and a good internet connection! Climate change will make life more unpredictable for everyone, but smallholder farmers will be particularly affected since a good harvest already depends on many different factors coming together in the right way and at the right time. For smallholder farmers to adapt successfully, they will need ready access to up-to-date information that can help them plan ahead. But several barriers currently make this far from straightforward. For instance, the various types of information they need - from market price data to soil maps - are typically spread across often very different kinds of databases, making it hard to integrate them into a single source. In addition, many of these datasets are not easily accessible or behind a paywall, and can also be difficult for the untrained to make sense of. Clearly this has to change. So, as part of the SFN Virtual Festival for COP26 , the network ran a four-day hackathon competition which took place between Tuesday 2 and Friday 5 November. Whilst global leaders discussed climate policy in Glasgow, cross-disciplinary teams of designers, coders, scientists, stakeholders and data analysts came together online to tackle the challenge set for them. This was to design an integrated open access platform for recommending crop type, planting date and management advice specifically for smallholder farmers in India. The team were provided with a range of suggested data sources, including agricultural market data from the Indian Government, farm surveys, World Soil databases from ISRIC , Remote Sensing datasets from the Google Earth Engine , and fertiliser, pesticide and disease data provided by SFN project partners. The timeframe was tight, particularly as the teams had to present their proposed platforms during a 10 minute ‘dragons-den’-style pitch to persuade a panel of judges to give them funding. “The primary factor we were assessing was the feasibility of the modelling approach, and the size and quality of the underlying data. But we also considered the technical expertise, motivation of the team, experience with local conditions, and adaptability” said judge Dr Angesh Anupam, a lecturer in data science at Cardiff Metropolitan University. Despite the entrants having just three and a half days to develop their platforms, the judges were so impressed by the designs that the first prize was jointly awarded to Samuel Bancroft and team Revive (led by Shamia Aftab). “The most impressive part of the two winning entries was their approach to modelling and their plans for scaling up the project if more real data became available. The synergy within team Revive also played a pivotal role” said Angesh. For Samuel, a PhD student at the University of Leeds, the Hackathon appealed to him because the challenge was closely related to his area of research: developing new machine learning methods to detect different crop types in satellite images. “Given my background, I used a machine learning approach to develop the crop type recommendation model, based on farmer survey data on what crops were planted in 2016-2020, alongside meteorological data and soil profile data. The resulting trained model was able to predict a crop type from data it had never seen before with over 50% overall accuracy. Whereas with just random guessing, the model would only be 5% accurate” he said. oth the winning entries successfully integrated multiple different kinds of data into a single portal that gave farmers a range of useful and clear recommendations. As Shamia, a Masters student in Data Science and Machine Learning at Peoples Education Society (PES) University (Bangalore, India), explained, an important part of the task was to consider the end-user’s experience and ensure the platform was fully accessible. “Our aim was to build a user interface that was simple and easy to navigate, without requiring high levels of literacy . So, for example, we used symbols and images as much as possible, and tagged locations to Google Maps. We also made sure our solution was scalable and could work on any cloud-based platform, so that it could be adapted to solve other purpose-driven problems.” Both Samuel and Revive will receive a £1500 cash prize and a package of software development support to enable them to improve their platforms further. “I was pleased with my model's performance and it is exciting to have the opportunity to continue working on it. I've already got a few ideas up my sleeve on how I can push the performance of the model even more” said Samuel. However even without winning, Samuel said it would still have been a valuable learning experience: “I particularly benefitted from having to develop the pitch for the judges at the end, from preparing my final presentation (right up to the last minute!) and putting together a concise set of slides, to thinking through my promotional spiel. It taught me a great deal about the process on how to best share my work with others, particularly in a way that showed my excitement about it.” Despite being an intense and daunting challenge, Shamia added that it was also highly rewarding and even fun. “There is absolutely no replacement for the thrill which we experienced and the marathon that our minds had to run in those three days to get something decent onto the table . Hackathons really help folks from all walks of academia and industry to come together and combine their rich knowledge base.” Samuel agreed: “I think hackathons are a great way to exercise the tenet of 'fail fast, fail often', where good (and bad) ideas can be realised in a short space of time. They can be competitive and intensive, but ultimately, a hackathon acts as a way to return to a childlike creative state, and explore different approaches in search of a stroke of serendipity.” You can learn more about our other events for the SFN Virtual Festival for COP26 on the dedicated webpage or on our summary blog post.
0 Comments
Your comment will be posted after it is approved.
Leave a Reply. |
AuthorJune 2024 - Archives
August 2024
Categories |