Amazon SageMaker customers for Geospatial ML

We use digital technologies like machine learning to explore the impact of heat on urban areas and the factors that influence local temperatures to deliver better design and support sustainable outcomes. Urban Heat Islands and the associated risks and discomforts are one of the biggest challenges cities are facing today. Using Amazon SageMaker geospatial ML capabilities helped us identify and measure urban heat factors with earth observation data, which significantly accelerated our ability to counsel clients. It enabled our engineering teams to carry out analytics that weren't possible previously by providing access to increased volumes, types, and analysis of larger datasets.

Rapid prototype development and continuous maintenance of our machine learning assets is one of our most important priorities. Amazon SageMaker geospatial ML capabilities provide us with a powerful solution to develop prototypes and govern our various models across their lifecycle. Using Amazon SageMaker, we were able to improve efficiency by almost 50%. Our teams were able to set up the required model pipelines and run testing in just 1–2 weeks, which is almost half of the time it took us previously.

When interviewing agriculture companies (e.g., seed, fertilizer, and chemical), we were surprised to find that almost all of them make multimillion-dollar decisions on limited, mostly subjective data. These companies often didn’t have the answers to the most critical questions, such as ‘How much land is available for seeding?’ or ‘When will the crop be ready for harvest?’ Being an avid user of various types of satellite imagery for almost 20 years, it was clear that we wanted to leverage the power of geospatial data to help agricultural companies uncover answers to these questions and help them make smarter decisions in the field. However, running complex machine learning algorithms on over 100 million acres of Australian cropland at 10m pixel resolution (and in real time) is extremely challenging. With Amazon SageMaker geospatial ML capabilities, we can go from concept to working model within a few hours.

Building scalable, state-of-the-art deep learning models that are cost effective is extremely challenging when dealing with large datasets from satellite imagery. Amazon SageMaker’s arsenal of tools helped simplify our end-to-end pipeline development of complex deep learning models. It saved us 25% in costs over the lifecycle of the model development, while also allowing us to focus on innovation solution development. We wanted to use Amazon SageMaker capabilities to develop a digital application that uses low-resolution satellite imagery for learning-based agriculture farm boundaries. With Amazon SageMaker’s built-in geospatial algorithms, Feature Store, and automatic parameter tuning, we were able to develop this application in an agile manner, ensuring we achieve all success metrics.

We wanted to use ML to help our customers increase agricultural yields with cost-effective soil maps. Amazon SageMaker geospatial ML capabilities allowed us to rapidly prototype algorithms with multiple data sources and reduce the amount of time between research and production API deployment to just a month. Thanks to Amazon SageMaker, we now have geospatial solutions for soil carbon sequestration deployed for farms and ranches across the US.