TGE’s goal is to speed up the development and commercialization of geospatial innovation. We focus on turning research-stage concepts into technology that is quickly usable by many organizations, creating new market opportunities. To do so, we identified a key technology gap in geospatial science—applying AI and Machine Learning to detect distinct features in satellite imagery. Opening this bottleneck could lead to even more groundbreaking innovations.
Innovation Bridge: The concept
We’re excited to share a recap of the 2024 Innovation Bridge program. The program builds a collaboration between academia and industry to reduce the time and funding barriers that often slow new concepts to reach commercial use. TGE connects academia and industry in four key ways: gathering stakeholder insights, ensuring the technology is commonly usable, making it cloud-accessible, and increasing awareness. After our first year running the Innovation Bridge, we’ve proven its effectiveness. Doing this within a nonprofit, rather than across individual companies, is not only cost-efficient but can also be a powerful driver of industry growth.
The industry side in this equation is represented by geospatial product companies, especially those using satellite imagery. These companies are eager to apply AI and Machine Learning to extract geographic features from the vast amounts of daily satellite data—reducing reliance on human analysis. These features power analytic tools for measuring everything from forest carbon to traffic patterns and global movements of people, planes, and ships. However, companies often lack the resources to identify, evaluate, and integrate new technologies effectively. Even when a promising technology is adopted, the time and cost of integration are often underestimated, leading to unexpected challenges and disappointing results.
On the other side of the effort, academic teams are primarily incentivized and regarded through publications and securing tenure, and many labs lack the resources to both publish their work and adapt it for industry needs. Academic teams also often have a limited view of industry applications.
Since commercialization isn’t a primary focus or funding priority in academia, we established a way to bring commercial concepts to the academic process. The Innovation Bridge provides funding to academic teams with breakthrough ideas, encouraging their collaboration with TGE without adjusting their primary research goals. This is done by bringing in industry experts to reflect industry needs and potential uses. TGE ensures these ideas are built into the innovation so they are practical and widely usable without changing the academic level of effort or focus. TGE funds technologists who work alongside researchers to refine, adapt, and publish their innovations, ensuring they are accessible and ready for broad industry adoption. As a result, product companies in the geospatial industry can evaluate the usability of new tech in less time, with less cost and decreased integration effort.
TGE’s Innovation Bridge initiative in 2024 focused on bringing a community of academics, technologists, and stakeholders together with a shared interest in establishing a collaborative approach to creating a global dataset of field boundaries. We have written extensively on the value of such a dataset in our blogs entitled “Field Boundaries for Agriculture Mid-point Progress Report”, “Celebrating Our Community’s Success”, and “Introducing Fields of The World”.
2024 Innovation Bridge: AI for Earth Observation and Field Boundaries
TGE kicked off the AI for Earth Observation and Field Boundaries effort in February 2024 (St. Louis Business Journal feature and TGE’s summary of the event).
TGE partnered with two world-class academics whose research advances the foundations and applications of machine learning for global data derived from satellite imagery: Dr. Hannah Kerner (Arizona State University) and Dr. Nathan Jacobs (Washington University in St. Louis). Dr. Kerner, Dr. Jacobs, and their respective labs were recipients of TGE Innovation Bridge grants which provided resources to orient their research outcomes toward commercial viability. To this end, TGE Industry Fellow Chris Holmes directed the strategy for following cloud-native geospatial principles in data schema development and data publishing. TGE Industry Fellow Jed Sundwall and the team at Radiant Earth provided guidance and the Source Cooperative infrastructure for open data sharing.
TGE brought in voices of stakeholders from across industry and government to exchange ideas that would help make outcomes more usable in the future. This committed group of contributors included Tristan Grupp, an Agricultural Data Scientist at the World Resources Institute, Jason Riopel, on the Ecosystems Services Micro Enterprises team at Bayer, and Ariel Zajdband, Senior Product Manager for Agriculture at Planet. As the project was progressing through the year, David McCaffrey, the Director of Remote Sensing at Miraterra Soil, and Ivor Bosloper, a technologist from the Netherlands with deep expertise in crop and farm field software solutions, both became aware of TGE’s work in this area and became committed collaborators and stakeholders in the effort.
In all, TGE supported the publication of the largest existing benchmark training dataset for extracting field boundaries from satellite imagery, called Fields of the World (FTW). Fields of the World was published as an academic paper that has been accepted for publication at the Association for the Advancement of Artificial Intelligence, a special track focused on Artificial Intelligence for Social Impact. While the paper was still in the peer-review process, anyone could view the data, examine the fiboa schema repository, explore tutorials for running the model, get instructions to further experiment, and access the data itself. There is also a history of more than a dozen blog posts describing the work as it progressed.
At the end of 2024, TGE held a team workshop followed by a public project showcase in St. Louis. The workshop included technical design sessions, core specification expansion and refinement, and planning for the upcoming phase of the effort in 2025. The Innovation Bridge Showcase featured a deep dive into the background of the issue at hand, a recap of the first phase of the initiative, and the plan for continuing the FTW project onwards.
2024 Innovation Bridge Impact
Less than a year from project launch, TGE and our partners were successful in ensuring that the academic advances using AI and ML to extract field boundaries from satellite imagery were immediately useful to a broad user base across industry. TGE facilitated:
- reduced time to market for commercial providers, governments, and non-profits by creating an open specification and open source tools for geospatial data,
- feedback loops between technologists and customers that increased the sustainability and scalability of the research outcomes,
- cost and time savings by relying on industry standards and utilizing existing cloud-native data-sharing platforms,
- decreasing the barrier of entry for new researchers, collaborators, and users via technical documentation and training materials, and
- increased awareness of the state of the art.
What’s Next?
We are excited to announce that we are kicking off Phase 2 of our work on AI for Earth Observation and Field Boundaries! We are hosting an in-person kickoff and sprint March 17-20, 2025 in St. Louis. TGE is currently seeking interested technical contributors as well as stakeholders from organizations who are interested in influencing the direction of the outcomes. If your organization is interested in either of these roles, please reach out to Jen Marcus (jen@tgengine.org) to discuss.