Unlocking Geospatial Insights — KSA Webinar Series
Article by: Charles Mwangi- Kenya Space Agency & Louisa Nakanuku-Diggs- Radiant Earth
Unlock Geospatial Insights: Sharing Knowledge from the Webinar Series
Geospatial data is becoming a crucial tool for professionals working in diverse sectors such as agriculture, natural resource management, disaster monitoring and environmental monitoring. Its efficient utilization is essential for unlocking valuable insights in policy formulation and enriching the decision-making process. The uptake and utilization of Geospatial data has the potential of acting as a catalyst for innovation, creating opportunities and empowering developing countries, like Kenya, to address pressing challenges and participate in the global space ecosystem.
Emerging AI application in EO
Recent trends have seen the combination of Earth Observation (EO) and Artificial Intelligence (AI) technologies to provide insights into numerous thematic fields of application of EO data. Currently, there are limited data sets available and easily accessible in Africa for use in training Machine Learning (ML) and Artificial Intelligence (AI) models. Most of the ML/AI models are trained on data from outside Africa, that carry different patterns or signatures than those from Africa. Running those models using data from Africa thus results in generating biased and inaccurate products.
Recognizing the importance of having easily accessible AI/ML training data, Radiant Earth, PLACE, Spatial Collective, and Kenya Space Agency (KSA), embarked on a collaborative pilot project using a drone to collect and curate labeled high-resolution agricultural data for open access and public benefit. Equally important, was the need to build technical capacity for Geospatial professionals in this emerging field through the webinar series that documents the process for ease of adoption and replicability.
A multifaceted project with a focus on knowledge sharing
The project had three phases: data collection, image labeling, and capacity building. Spatial Collective spearheaded the data collection and image processing aspects, working closely with PLACE and KSA. They documented the practical steps taken during the implementation process, encompassing various elements, such as image, drone, and flight requirements. Radiant Earth and Spatial Collective collaborated on the image labeling phase, training a small team of labelers from KSA to identify building and field outlines. The labeled drone data can be found on this link.
KSA was keen to utilize the project to build their technical capacity and actively participated in the project, gaining valuable knowledge and skills from the project partners. This knowledge transfer included hands-on training in field data collection techniques and image labeling methods. One of the key objectives of the project was empowering KSA to independently undertake similar activities in the future.
Unveiling the Knowledge Sharing Webinars
Appreciating the need for capacity development for Geospatial practitioners in Kenya and the region, the project incorporated a component on capacity building through a THREE-part webinar series. The webinars were designed to disseminate the knowledge and techniques developed during the project’s implementation.
While the live webinars concluded, the recordings remain a valuable resource for anyone seeking to learn more about process, the geospatial data used and application of the machine learning models. Access the recorded webinars to learn more about the practical applications of these tools in addressing real-world challenges. The Webinar series was moderated by Charles Mwangi, Acting Director, Space Sector and Technology Development at, Kenyan Space Agency.
Webinar 1: SpatioTemporal Asset Catalog (STAC)
The first webinar, titled “SpatioTemporal Asset Catalog (STAC),” featured Michelle Roby, Developer Advocate at Radiant Earth. Michelle introduced participants to the STAC specification, which provides a standardized way to organize and describe data. The STAC enhances geospatial data accessibility and user-friendliness, making it easier to discover, use, and share geospatial information.
View the recording here and access the presentation slides.
For additional resources on STAC click on this link
Webinar 2: Field Boundary Detection Model
Daniel Nwaeze, Data Scientist at Radiant Earth, led the second webinar, diving into the intricacies of transfer learning using the South Africa Field Boundary Detection Model. This hands-on session showcased how to adapt this open model to identify agricultural fields in the Murang’a region, located on the outskirts of Nairobi, Kenya. Participants gained insights into the practical application of transfer learning in the field of geospatial analysis.
View the recording here and access the presentation slides.
View Github repository on this link
Webinar 3: Building Footprint Segmentation Model
James Haithcoat, Principal Consultant at DevGlobal, guided participants through the process of adapting the Ramp Baseline Model for Building Footprint Segmentation. This webinar focused on the practical applications of the project’s outputs, highlighting the model’s use in urban and infrastructural planning.
View the recording here and access the presentation slides.
View Github repository on this link
Empowering the future through knowledge and data
The project’s overarching goal of democratizing access to high-resolution agricultural data for public benefit is a critical step toward Kenya’s development. By reviewing the recorded webinars, you can equip yourself with the knowledge and skills necessary to contribute to this mission. The webinars and the shared links are a great entry point for learning about geospatial technologies, models, and applications.
