SE15-ICT

 

[SE15-ICT] Data-Driven Farming for Everyone
 
DAY 2 – Thursday 28 – 14:20-15:50
Swiss Tech | Room 1A | Level Garden 
 
Session Leaders
Daniel Jiménez R.
International Center for Tropical Agriculture (CIAT)
d.jimenez@cgiar.org
 
Daniel Jiménez is a scientist at the International Center for Tropical Agriculture (CIAT). He specializes in using data mining techniques to help decision-makers enhance and accelerate the impact of agricultural research to address pressing challenges such as yield gaps and climate change. His work has been twice recognized by the United Nations for using big data (2014) and information and communications technologies to address climate change (2017) and was selected as one of the winners of the first World Bank Group Big Data Innovation Challenge (2015). A holder of PhD in Applied Agriculture from Ghent University, he was previously affiliated with Bioversity International, the University of Applied Sciences of Western Switzerland (HEIG-VD) and the French Agricultural Research Centre for International Development (Cirad).
 
Andrea Gardeazabal Monsalve
Centro International de Mejoramiento de Maíz y Trigo (CIMMYT), Mexico
a.gardeazabal@cgiar.org
 
Andrea works for CIMMYT’s Sustainable Intensification Program, on data-driven agronomy, knowledge management and ICT as vehicles for innovation in agri-food systems. She holds an MSc in Information and Communication Technologies for Development at the University of Manchester, UK, and an MSc in Political Science from the Los Andes University, Colombia. Andrea has over a decade of experience designing and deploying large ICT for agriculture and education projects in Mexico, Guatemala, Colombia and Ecuador; she has been also guest speaker at several ICT and Big Data Conferences. Andrea currently coordinates the Monitoring, Evaluation, Accountability and Learning Unit for Latin American projects at CIMMYT, which entails the design and operation of robust information systems for the collection, analysis and dissemination of data, and co-leads the MasAgro Project that pioneers innovation hubs with participative research approach in the region. She prefers and fosters open-source systems, and she is interested in implementing solutions using the Blockchain technology.
 
Summary
 
A wide range of services, mostly in developed countries; offer farmers the possibility of make better data-driven decisions including what to plant the next season, most profitable crop, right amount of fertilizers, etc., These services are currently used by farmers that are early adopters of technologies and that can afford such services, what about the other farmers? Developing countries? How can we contribute to reduce such inequity? How should the academy, for-profit and non-profit organizations work together?
 
 
Panelists and Abstracts
 
Deep Learning for Surveillance of Fall Armyworm Infestation
Amanda Ramcharan1, Ravishankar Narayana1, Peter McCloskey1, David Hughes2
1Department of Entomology, College of Agricultural Sciences, Penn State University, State College, PA, United States
2Center for Infectious Disease Dynamics, Huck Institutes of Life Sciences, Penn State University, State College, PA, United States
 
Presenting author’s email address: amr418@psu.edu
 
Biography of Presenting Author: Dr. Ramcharan researches artificial intelligence in agriculture. A native of Trinidad and Tobago, she received a BSE in Mechanical Engineering from Princeton University, then went on to a Princeton in Africa Fellowship in Kenya for one year. After this, she completed a PhD in Agricultural Engineering with a minor in Computational Science focused on applying machine learning algorithms to soil mapping. Her current work with PlantVillage focuses on applying the latest computer vision models to detect plant diseases through smartphones.
 
Abstract: The recent discovery of fall armyworm (FAW) in 2016 in Africa is considered to be a huge threat to food security in the region. Fall armyworm is a major problem for important crops such as maize and novel methods of FAW detection are needed to support surveillance and report damage. Object detection, a type of deep learning model, is a scalable technology to observe and report FAW damage and Google’s TensorFlow platform offers an avenue for this technology to be easily deployed on mobile devices. Using a crowdsourced dataset of FAW damage on maize, we trained an object detection model to identify FAW leaf damage, frass, and eggs. The best trained model was deployed on an Android device and detected FAW categories in the field for maize plants never seen by the model. These results show that object detection offers a fast, affordable, and easily deployable strategy for pest surveillance.
 
IoT and AI Methods for Plant Disease Detection in Developing Countries
Sandor Markon1, Hadjer Hamaidi1, Tun Tun Win1, Ronald Criollo2, Ryo Ohtera1, Marco Zennaro1,3
1Kobe Institute of Computing, Kobe, Japan
2Escuela Superior Politécnica del Litoral, Ecuador
3Abdus Salam International Centre for Theoretical Physics, Trieste, Italy
 
Presenting author’s email address: markon@kic.ac.jp
 
Biography of Presenting Author: S. Markon received his PhD. from Kyoto University (Japan), during his tenure at the Japanese company Fujitec as head of the R&D group, developing embedded systems, AI-based control methods and other computer applications. Currently he is a professor at Kobe Institute of Computing, engaged in research and education concerning IoT, human-machine interfaces, optics, optimization, embedded systems and other topics. Dr. Markon is a director at the Institute of Complex Medical Engineering and at 3 commercial companies.
 
Abstract: Plant disease detection and prevention is a major issue for countries that are dependent on agriculture, but have only limited resources for advanced agronomical techniques. To help farmers, we are developing a combined disease detection technology, using low-cost sensors, cameras, and telecommunication. The images from cameras can identify plants infected with insects, fungus or viruses; while the physical sensor data can help recognizing environmental conditions indicating the danger of infections. We also intend to use hyper-spectral imaging techniques to calibrate the color recognition of off-shelf cameras in order to improve selectivity and sensitivity. Currently we are focusing on date palms in Algeria, rice in Myanmar, and banana in Ecuador, but we believe that our techniques should be useful in many other agricultural areas.
Eventually we intend to contribute to the PlantVillage project with curated versions of our datasets and methodology.
 
MyWell: Crowdsourcing Water Data for Sustainable Farming
Lewis Daly1, Basant Maheshwari2
1Vessels Tech, Adelaide, Australia
2Western Sydney University, Syndey, Australia
 
Presenting author’s email address: lewis@vesselstech.com
 
Biography of Presenting Author: Lewis Daly is the Founder and CTO of Vessels Tech in Australia. He has a background in IT and Business, with a Masters degree from Carnegie Mellon University in Adelaide. He has also worked as an IT Systems research intern at Hitachi in Japan.
 

Abstract: Across India, over 60% of agriculture relies on groundwater. Groundwater is often over-exploited, and groundwater depletion threatens many livelihoods. Existing interventions have been insufficient, and often fail to engage farmers at the lowest level. MyWell sets out to build a participatory, bottom-up approach to groundwater management. MyWell is an application for data-driven insights of groundwater at the village level. MyWell crowdsources groundwater indicators from a group of connected farmers using SMS or a smartphone application. In this paper, we show how MyWell has the potential to impact the lives of farmers by (1) empowering them to participate in science, and finding solutions to groundwater scarcity, and (2) assist them in building communities which work together to conserve groundwater. Then, we assess MyWell’s application across two watersheds in rural India. We show how farmers are using MyWell to gain visibility into the groundwater situation, and examine the limitations of MyWell’s approach.