| Description |
The CoSIA dataset (Land Cover by Artificial Intelligence) was produced and used within the framework of the French large-scale reference land cover and land use mapping product (OCS GE). This database predicts land cover (that is, the nature of the soil) according to 15 classes:
· Building
· Impermeable area, Permeable area
· Swimming pool, Water surface, Snow
· Greenhouse, Bare soil
· Coniferous, Deciduous, Shrubland, Lawn, Crop, Plowed land, Vineyard
Main characteristics:
These data are raw and obtained through artificial intelligence (AI) processes, notably deep learning methods. They represent so-called "prediction" information. They are obtained from a digital AI model that statistically estimates for each pixel its membership to a class. These data therefore do not comprehensively reflect the reality on the ground: There are margins of error that are documented for each object class in the links provided below. |