Wherobots, the Spatial Intelligence Cloud founded by the original creators of Apache Sedona, has officially announced the general availability of Raster Inference for WherobotsAI.
According to certain reports, Raster Inference arrives on the scene with an ability to turn satellite or drone imagery analytically accessible for data developers who use SQL or Python. More on the same would reveal how it features flexible pay-as-you-go pricing and support for Wherobots hosted or custom computer vision models. As a result, the technology allows for its capabilities to be built from aerial imagery in hours versus weeks or months, and that too, at a fraction of the cost associated with starting from scratch.
To understand the significance of such a development, we must take into account how satellite and drone imagery are already fueling advancements across defense, agriculture, energy, transportation, climate science, clean-tech, and insurance. Having said so, they also make up these massive, unstructured datasets that are complex and multidimensional, whereas on the other hand, the go-to computer vision tools used to extract insights from this imagery are not designed for the scale and variety of this data.
Not just that, comprehensive experience is also needed to set up and manage specialized infrastructure, tune it to improve model performance, and shuttle inference results, the boundary polygons and coordinates that identify features of interest like buildings, flood damage, or crops, into a secondary solution capable of working with this data. Even with that, though, it can take weeks or months for teams to stitch together what are often fragile solutions, causing costs to rise dramatically and making solutions from this data hard to attain.
In response, WherobotsAI Raster Inference enables the creation of solutions based on satellite and drone imagery to be significantly easier, lower cost, and remain accessible to existing data teams.
Talk about the whole value proposition on a slightly deeper level, we begin from its on-demand Inference. This translates to how the solution delivers production ready inference pipelines that, on their part, can be created in seconds, on-demand, with no infrastructure to manage. Next up, we must get into its model integration, which allows you to start quickly with Wherobots’ hosted machine learning models or easily import existing models.
“There are very few model inference products built for the unique needs of geospatial workloads, yet there are countless use cases that can benefit from existing models, and aerial imagery,” said Jia Yu, co-founder and Chief Architect of Wherobots. “We co-developed the MLM STAC extension to make models useful between organizations, use cases, and across platforms. And Wherobots Raster Inference does all the heavy lifting required to run these models in the background.”
Then, there is the prospect of accessing a unified environment, where you can easily join inference results with other datasets using Spatial SQL and Python. This can be done through a spatial engine which is up to 20x more performant than popular alternatives. Another detail worth a mention here is rooted in the technology’s workload automation capacity. The stated capacity, you see, treads up a long distance to facilitate job runs and integration, while Apache Airflow simultaneously make it easy to automate insights.
Hold on, we still have a few bits left to unpack, considering we haven’t yet touched upon the solution’s promise to provide improved model portability. Leveraging support for the MLM STAC (Machine Learning Model Spatial Temporal Asset Catalog) extension, users can access a new standard co-developed by Wherobots to enhance model portability across platforms.
Complementing the same is how Wherobots also provides you with example notebooks for Wherobots hosted models to help customers get started.
“Every day, petabytes of satellite data are produced. But without specialized talent, deep pockets, or significant time investments, this data often remains untapped, putting solutions out of reach,” said Mo Sarwat, co-founder and CEO of Wherobots. “Raster Inference changes the game by enabling teams to easily derive actionable insights from aerial imagery, on-demand. This solution puts unprecedented power into the hands of developers and data scientists.”
