Datashader is a graphics pipeline system for creating meaningful representations of large datasets quickly and flexibly. Datashader breaks the creation of images into a series of explicit steps that allow computations to be done on intermediate representations. This approach allows accurate and effective visualizations to be produced automatically without trial-and-error parameter tuning, and also makes it simple for data scientists to focus on particular data and relationships of interest in a principled way.
via http://datashader.org/