Foundation model

A large model pre-trained on vast unlabeled Earth-observation data, then fine-tuned for specific tasks (flood mapping, crop type, cloud gap-fill) with relatively little labeled data.

Why it matters

Foundation models (e.g. Prithvi, Surya) let small teams reach strong performance without training from scratch — but the weights are open, not turn-key, so a fine-tuning step is usually required.

In plain terms

A model that already “knows what Earth looks like” from huge pre-training, so you only have to teach it the last mile for your specific question.