Why OversightML Chose Cesium

4 minute read
Content level: Foundational
2

This article discusses the integration of two open-source projects: OversightML (OSML), which is part of AWS Open Source Guidance, and Cesium, a 3D mapping platform. Both OSML and Cesium are designed to support geospatial and imagery workloads, making them complementary tools in the field of geospatial application development.

In the dynamic landscape of overhead imagery processing, analysis, and storage, visualization is a crucial component for efficiency and insight. With OversightML (OSML), we strive to provide state-of-the-art open source solutions to empower customers to adopt the cloud for geospatial application development, which is why we chose the Cesium platform to build a geospatial visualization tool to improve our developer and customer experience.

Motivation: Bridging the Visualization Gap When we first started, OSML required manual steps for visualizing source imagery and model detections processed by ModelRunner, the component of OSML that performs inference on overhead imagery. This involved submitting hand-coded image processing requests to OSML, downloading imagery and detection data from S3, and importing them into other open-source GIS tools for visualization and analysis. To empower our developers and users by streamlining these processes, we integrated OSML and Cesium.

Performance: On Demand Rendering of Satellite Imagery and Terrain Cesium stands out for its performance, being built on WebGL, which allows it to render complex 3D graphics quickly and smoothly. This capability is vital for OversightML, where real-time ingestion of extensive and multi-variate datasets is a necessity. The performance of Cesium ensures that our visualizations run efficiently, even with the high demands of processing and displaying large overhead imagery data sources such as NITF and GeoTIFF files.

Cesium also supports 3D Tiles, an open standard created by Cesium for streaming massive 3D geospatial datasets. By transforming terrain datasets into 3D Tiles, we can host and display them in Cesium, enabling users to verify outputs against real-world features. These tiles are then hosted locally and served to the Cesium client, providing a seamless and integrated visualization experience. This architecture ensures that our visualization tools are not only powerful but also user-friendly and accessible, meeting the diverse needs of our users.

Flexibility: Customization and Integration with OSML The flexibility of Cesium is another significant advantage. With support for 3D terrain, overlay widgets, and various data sources, our developers can create a highly customized and interactive experience to interact with OSML. As an open-source project, Cesium benefits from ongoing maintenance and upgrades, ensuring we have access to the latest features and improvements without additional licensing costs. This flexibility extends to supporting isolated environments, which is crucial for certain use cases within OSML.

A key aspect of our solution is the ability to display GeoJSON FeatureCollection objects returned by our applications. Cesium enables us to generate and visualize these features directly within the globe, including metadata, bounding boxes, and object masks. This direct integration simplifies the process for developers and operators, allowing them to generate inputs and inspect outputs for OSML without additional tools or configurations.

A Better OversightML Experience By choosing Cesium, we have enhanced OversightML's visualization capabilities, providing our developers and users with a powerful, flexible, and scalable tool for geospatial analysis. The integration with ModelRunner workflows, support for large datasets, and advanced 3D capabilities ensure that we continue to deliver exceptional solutions in the geospatial domain. We are excited about the future possibilities this integration brings and remain committed to innovation and excellence in overhead imagery processing.

For those interested in diving deeper into OversightML and its capabilities for processing overhead imagery on AWS, we encourage you to visit our official AWS guidance page. This comprehensive resource provides detailed information about the architecture, implementation, and best practices for using OversightML.

To explore further, please visit: https://aws.amazon.com/solutions/guidance/processing-overhead-imagery-on-aws/