Repository for the Machine Learning for Earth Observation Summer School in Pavia, Italy
Setup the conda environment locally with:
conda env create -f conda-env.yml
Or use mamba instead (identical result with faster dependency solving):
mamba env create -f conda-env.yml
This will create an environment named bip-ml4eo
. Activate with:
conda activate bip-ml4eo
Now you can run the code and develop in this repo (except for non-Python code).
The Sentinel-2 data has been uploaded to Kaggle: see sentinel-2-cremona dataset. The ancillary data (hydrometric levels) is on Google Drive here.
The data should be downloaded at data/Water/
with the folder tree as below:
data/Water
├── Cremona SIAP - Livello Idrometrico - 2023-01-12.csv
└── sentinel-2-cremona
├── images
│ ├── S2A_MSIL2A_20220309T100841_N0400_R022_T32TNQ_20220309T134626.tif
│ ├── S2A_MSIL2A_20220309T100841_N0400_R022_T32TNR_20220309T134626.tif
│ ├── S2A_MSIL2A_20220322T101711_N0400_R065_T32TNQ_20220322T141030.tif
│ ├── S2A_MSIL2A_20220322T101711_N0400_R065_T32TNR_20220322T141030.tif
│ ├── ...
└── S2_Cremona_DESCENDING_info.csv
Merge the Sentinel-2 raster tiles with:
python src/merge_s2_tiles.py
This should take approx. 3-4 min to complete.
This should create 31 merged images in total. The merge dataset will be in the output directory data/Water/sentinel-2-cremona/merged-images
.
This notebook contains Olivèr's work.