Visualizing analytic raster data in the browser still typically depends on backend services that render pre-computed raster tiles (e.g. PNG/JPEG) and frontend layers that are reliant on those services. That approach works — but it adds infrastructure, security overhead, and limits how people can change rendering on demand.

With our work on deck.gl-raster, we’re bringing high-performance, extensible raster rendering to the deck.gl geospatial visualization library. This adds to our existing work speeding up visualization of large geospatial vector data — with the goal of making it realistic to render both massive vectors and analytic rasters in the same interactive view.

This talk dives into the technical bits of why client-side raster visualization is challenging and explains our implementation. In the process, we’ll learn about raster reprojection, how GPUs render image data, and how our reprojection implementation leverages mesh-based rendering for efficient and accurate reprojection.

This talk will be decently technical, but endeavors to be accessible to anyone familiar with raster data like Cloud-Optimized GeoTIFFs.