The Generic ML pipeline of EO4EU enables the learning of a robust and transferable representation of the input data in a latent space, in an unsupervised way.
Minimize the requirement of use-case specific labelled data.
Rely on approaches such as contrastive learning of visual representations. Such a representation will provide the input representation over which the downstream tasks will operate for the learning of task specific models.
It allows the owners of the downstream tasks and the related use cases to easily train such task-specific models with a minimum data labeling effort.