Published December 18, 2023
| Version 1.0.0
Dataset
Open
CPAZMAL: Cryosphere PAZ satellite MAchine Learning
- 1. LISTIC, University Savoie Mont Blanc
- 2. Université Savoie Mont-Blanc
Description
CPAZMAL: Cryosphere PAZ satellite MAchine Learning
The aim of this dataset is to serve as a foundation for machine learning in multi-class classification, specifically in mountainous regions. It comprises descending images acquired by the PAZ X-band satellite, focusing on the Mont Blanc region during the period from January 2020 to November 2021, totaling 56 acquisitions.
The time series is divided into two sub-sections:
- From January 2020 to 8th January 2021 included: dual polarisation HH and HV,
- After 8th January 2021: single polarisation HH.
The datas are divided into 8 classes:
- Hanging Glacier (HAG)
- Ice Aperon (ICA)
- Ablation area
- Accumulation area
- Rock
- Plain
- Forest
- City
In each classe, between 4 to 10 groups or distinct areas, where their complete description (position, aspect, elevation, ...) can be found in the desc_topo_areas.png file
We provide code that directly extracts temporal or spatial datasets, consisting of homogeneous windows paired with respective labels.
# Request and save data into hdf5 file
rqtemp = "classe in ['ICA','HAG','ABL','ACC','FOR','CIT','ROC','PLA'] & date < '2021-01-01'"
cdlf = Dataset_tiff2hdf5
(
path_to_folder_extracted,
different_group=True,
n_jobs=1,
outpath="path_to_dataset.h5",
extension="temporal"
)
cdlf.extract_data(rqtemp, polarisation="HH", winsize=7, save=True)
# Load the previously extracted data set
(
x,
y,
gr,
org,
_,
) = load_h5(path_to_dataset.h5)An example of how to use it can be found at Github.
The authors would like to thank the Spanish Instituto Nacional de Tecnica Aerospacial (INTA) for the PAZ images (Project AO-001-051)
Files
desc_topo_areas.png
Additional details
Related works
- Is continued by
- Computational notebook: https://github.com/Matthieu-Gallet/PAZ_DTW_classification (URL)