PIXSTART (FR)
This project developed a service for automated clustering of Sentinel-2 pixels which allows its users to focus on Earth surface changes rather than on remote sensing problems, and hence to develop their own downstream applications.
Available via a standard interface, it produces on demand and in near real time classified Sentinel-2 images at 10 meter resolution, in a new 2D space which preserves all properties of Sentinel-2 information (spectral complexity and topology, maximum spatial resolution), and which is consistent over time and space (from a Sentinel 2-tile to another).
This simplified new space is allowing pixel labelling or grouping by a posteriori classes identification, change detection by distance computation, interpolation, and may be used as a pre-processing step for all kinds of machine learning algorithms.
