AI-based compression of satellite data and images - Ref. 94/26/IR

The project deals with AI-based compression of satellite data and images. The compression process can use an autoencoder-based approach or a prediction-based approach; the former is more suitable for high-compression applications, while the latter is well suited to high-quality applications. The project specifically addresses AI-based models performing compresison directly onboard the satellites. This requires to develop models that are compatible with the memory and computation constraints of computing platforms available for satellites, and requires to adapt the design to deployment on hardware suitable for embedded systems.