Summary
Exploitation and analysis of heterogeneous data are usually conducted thanks to specific methods, generally dedicated to each kind of data, which account for the measurement process and the very nature of the data itself. Within advanced scenarii of concomitant availability of several yet distinct measurements, fully analyzing these extended datasets requires a unifying framework overcoming a crude and marginal description of a single measurement.
The main objective of this research program is to develop learning algorithms able to extract meaningful information from multi-source, multi-scale and multi-temporal data. Particular applicative contexts deal with remote sensing for Earth observation and automotive systems.
This project is part of the ANITI Integrative Program "Acceptable AI" (IP-A)
and contributes to the themes "Learning with little or complex data"
and "AI and physical models".
