Taylor Geospatial Engine’s First Innovation Bridge

Apr 3, 2024

Following our last blog post outlining Taylor Geospatial Engine’s (TGE) Innovation Bridge program, we are pleased to now share details about our first initiative. As a reminder, the Innovation Bridge Program is intended to facilitate and accelerate the process of commercializing geospatially-centered research and development projects and prototypes. Our process framework is centered around engaging a broad set of experts and stakeholders to:

  • Align practitioners on best practices, industry standards, and shared expertise
  • Facilitate collaborative technical development to address common gaps and deficiencies in the geospatial industry
  • Create talent pipelines and user feedback loops that better connect the research community to industry demand
  • Advance technical roadmaps in high demand industry areas within the research and development ecosystem.

The Innovation Bridge framework creates opportunities to advance geospatial research and development efforts for commercial use cases via relatively short and highly targeted initiatives which each last approximately nine months.

INITIATIVE NUMBER ONE: INNOVATING WITH AI, EARTH OBSERVATION IMAGERY, AND THE CLOUD

Our first initiative, launched in February 2024, focuses on advancing the state of the art in the application of AI and computer vision to earth observation imagery.  At the same time, the project is promoting the use of cloud-native techniques and collaborative data platforms to enable the sharing of high-value global geospatial data products across a broad community of consumers. Taylor Geospatial is thrilled to be partnering with and utilizing Cloud-Native Geospatial Foundation tools and Source Cooperative for this first initiative.

Commercial markets have traditionally underinvested in analytics and insights derived from satellite imagery (aka earth observations) even though the storage and processing power is available in the cloud, and large-scale data analysis using AI has made rapid progress. One reason for the underinvestment is the complexity of the task with too few human experts: the amount of data is staggering; the number of features that could be extracted from satellite imagery is huge; the expertise at applying AI and CV to satellite imagery is limited by the number of trained experts in the world – and that limited expertise is held in a few major companies (Google, MSFT, Meta, etc) or distributed across many individual research labs in universities across the globe.  

This is an opportunity to apply the Innovation Bridge framework to draw expertise and approaches from the research arena and apply them to the problem while expanding the body of expertise.  Future commercial potential comes directly from the technical collaboration around these problems.

TGE SEED FUNDING: MACHINE LEARNING AND COMPUTER VISION APPLIED TO GEOSPATIAL DATA

TGE is providing seed funding to two outstanding research teams (Dr Hannah Kerner and Dr Nathan Jacobs) to extend AI models for extracting field boundaries from satellite imagery, benchmark those models towards accuracy necessary for commercial use, establish an approach to a global field boundaries dataset schema, and use common open platforms for sharing the foundational dataset. This is an initial step to enable commercial innovation and adoption of satellite-derived data for Ag Tech and Food Security applications. 

COLLABORATORS: INDUSTRY, GOVERNMENT, NGOs

In order to ensure academic capabilities are relevant to industry issues, TGE worked diligently to bring together a community of technologists from mainstream IT companies, government, NGOs, and the Ag Tech markets for this collaborative initiative.  To define plans for the joint effort, we gathered in person at Washington University in St. Louis for a two-day workshop in late February.  The workshop was just the start – the participants laid the foundation for the project which will continue for the next nine months in sustained development. At the end of the workshop, the community had catalyzed and was oriented around key technical topics and some pre-defined measures of success.

Seeing the willingness to collaborate and dig into existing technical challenges reinforced our hypothesis that the time for this initiative is now. There’s been a lot of great academic work on field boundaries, many agricultural companies have tried many different approaches and are seeing the potential of AI/CV to help, there’s a growing number of companies selling field boundaries and related data commercially, there’s a number of national field boundary datasets, of varying quality, and there’s a few efforts to run the AI/ML algorithms at scale with compelling results. The group of organizations that have joined us is very compelling.

The following organizations participated in the workshop in St. Louis:

  • Varda – An initiative building collaboration around FieldID – a global system for identifying any field.
  • Microsoft AI for Good – A research group that has been building innovative global scale data products from satellite imagery using AI/ML and is working on field boundaries.
  • Danforth Plant Science Center – A non-profit that brings together the best and brightest plant scientists from around the world to answer some of humanity’s most profound challenges.
  • World Resources Institute – The non-profit behind Global Forest Watch and a number of other sustainability initiatives, who is starting work on field boundaries.
  • Oak Ridge National Lab GeoAI Group – One of the most innovative geospatial organizations in the national government, bringing deep experience with AI/ML..
  • Radiant Earth – The non-profit behind Cloud Native Geospatial Foundation and Source Cooperative, who has done extensive work on agricultural ML models.
  • Planet – Provider of earth observation imagery as well as higher-level data products for agriculture including Field Boundaries, Crop Biomass, Soil Water Content, and Land Surface Temperature.
  • Bayer – One of the largest agricultural companies in the world, represented by a group working on sustainable agriculture and carbon.
  • Intent Ag – A rapidly growing Ag Tech company focused on field trials, business decisions and sustainability
  • TetraTech – A consulting firm that is implementing the Enabling Crop Analytics at Scale (ECAAS) initiative, funded by the Bill & Melinda Gates Foundation, that aims to catalyze the development, availability, and uptake of agricultural remote-sensing data and subsequent applications in smallholder farming systems. 

Select participants were not able to travel, but provided key input in two remote sessions during the workshop:

  • Google Research– Representatives from the sustainable agriculture team shared their work on the Agriculture Understanding Platform that is creating field boundaries and other ML-derived insights  for all of India.
  • DigiFarm – A startup building field boundaries and other agriculture insights with AI/ML and satellite imagery.
  • Regrow – An Ag Tech company focused on Resilient Agriculture and detecting sustainability practices, with extensive experience using AI/ML to create field boundaries from imagery.
  • UN-FAO – Leading the EUDR forest regulations, that will require companies to prove the fields they’ve sourced from have not been recently deforested.
  • Open Supply Hub – A non-profit seeking to map and create a unique identifier for every supply chain production point in the world, whose work will touch field boundaries in the future.
  • ALCIS – An organization working on using geospatial insights to help the sustainable development goals.
  • Fall Line Capital – An innovative investor in farmland and Ag Tech companies. 

It was inspiring to see so many diverse interests come together to work on a common cause. In the next blog post we’ll dive deep into what was accomplished at the initial workshop and what is planned for the rest of the initiative. We’ll shift to working fully remotely, in an open source style that anyone can join in.

WHAT’S NEXT

Technology innovation will reach consumer markets faster when a bridge exists for academia, the commercial industry and government to share best practices, validate and scale AI models for common problems, and publish data on an open cloud. Taylor Geospatial Engine is excited to be building a framework and providing the ecosystem with tools and platforms to make this happen.  Please stay tuned as we progress through this first initiative and launch additional initiatives and opportunities to collaborate.