Rapid Sen2Cor is a platform that enables better handling of Earth observation data. It uses deep learning models to accelerate the process of bottom-of-atmosphere (BoA) correction, which is a necessary step in remote sensing. This approach can convert a Sentinel-2 tile (Level-1C) into a useful tile (Level-2A) in less than 10 seconds instead of the normal 30 minutes. Along with time, there is also a significant reduction in the energy utilised by the hardware for data pre-processing.
This use of deep learning addresses the current problem in scaling Earth observation data, which involves simply buying more hardware. The computational requirements for hosting and correcting data are also directly responsible for many end-user limitations, such as rate limits and data availability. This approach can help combat these limitations while empowering new ways of handling Earth observation data.
The initial prototype for Sen2Cor exhibits higher than 99% accuracy for Sentinel-2. This approach to rapid BoA data can work with many different algorithms, which is why Sen2Cor will also be expanded to accelerate broader range of them.

ClearSky Vision
Morten Fjord Pedersen, Malthe Jensen, Kaare Madsen