search

ecmwf-projects / ai-vegetation-fuel

Predicting Fuel Load from earth observation data using Machine Learning
https://ml-fuel.readthedocs.io/en/latest/
Other
14 stars 5 forks source link

Consider removing resampling step #13

Open cvitolo opened 3 years ago

cvitolo commented 3 years ago

Raw features are in a gridded format (either GRIB or NetCDF) and are resampled to a common spatial resolution in the pre-processing step. After resampling, features are extracted at each point of interest. The machine learning models implemented use in input a table where features and outcomes are columns and each row correspond to a point (lon and lat coordinates).

Please note the resampling step introduces errors/loss of information and it's not necessarily needed in this case. Please consider inspecting each raw feature in its original spatial resolution, to minimise information loss.

cvitolo commented 3 years ago

This issue is solved in branch: cvitolo-patch-1. Training script will need some modification to accept new inputs. Will close this issue when the branch is merged into master.