Ryan Avery developing ML system for Omidyar Network at Clark Labs

  • Funding
  • News
  • Research tags:
  • Smallholder Agriculture
  • Ghana
  • Remote Sensing
  • Deep Learning
  • Ryan Avery excerpt: "Ryan is working as a Researcher at Clark Labs to develop the scalable machine learning component of an active learning system to detect smallholder agriculture"

Ryan Avery will be at Clark Labs developing software for the project "Developing and Scaling Up The Mapping Africa Active Learning Platform", funded by the Omidyar Network. He will be working with the team to build a scalable system for detecting smallholder farm fields across Africa, focusing first on a sub-region of Ghana. Crowdsourced maps of smallholder fields have been shown to be an effective data source to train models to predict the locations of unmapped fields. Ryan work this summer will be to implement neural network and random forest models on Amazon Web Service's cloud computing infrastructure to locate fields with near human-level ability at flexible spatial scale. Ultimately the deliverable sof this project will be a rich dataset of smallholder agriculture locations across Africa, opening new possibilities for targeted research on changing land use patterns and land cover in dynamic agricultural regions. The methods work is a continuation of Debats et al. 2016 Debats et al. 2017. He will return to UCSB in mid-September to start his second year, and will continue to collaborate with the project team remotely.