Computational geoscientist, researcher in geoinformatics and Earth observation.
This account hosts the source code accompanying my scientific publications: scripts, data and figures for reproducible processing, analysis and cartographic visualisation of geospatial data. Repositories are named by method and study area and are linked to the corresponding peer-reviewed articles.
- Deep learning for Earth observation in PyTorch: spatio-temporal forecasting of land-cover time series (ConvLSTM, Transformer self-attention) and temporal CNNs for Sentinel-2 time-series classification
- Explainable machine learning for remote sensing: SHAP, gradient-based saliency, spatially grouped cross-validation and statistical model comparison
- Machine-learning classification for remote sensing (Random Forest, LightGBM and gradient boosting, SVM, MLP/ANN, Decision and Extra Trees)
- Satellite image processing and land-cover mapping from Sentinel-2 MSI and Landsat OLI/TIRS time series
- Object detection and machine-learning-ready datasets from satellite imagery: Hough-transform labelling, instance masks, spatial-block splits
- Cloud geoprocessing and reproducible pipelines: Google Earth Engine (Cloud Score+ compositing) to PyTorch, PySpark
- Marine geophysics, bathymetry and geomorphometry of ocean seafloor
- Cartographic scripting, terrain modelling, seismology and geohazard assessment
Python, R, JavaScript (Earth Engine), Octave/MATLAB, Shell (bash, AWK). PyTorch, scikit-learn, LightGBM, SHAP, scikit-image, rasterio, GDAL/PROJ, GMT and pyGMT, GRASS GIS, QGIS, Google Earth Engine, Apache Spark (PySpark), ObsPy, terra, earthpy, ggplot2, seaborn.
- ORCID: https://orcid.org/0000-0002-5759-1089
- HAL (CV and publications): https://cv.hal.science/polina-lemenkova
- ResearchGate: https://www.researchgate.net/profile/Polina-Lemenkova-2
- Scopus: https://www.scopus.com/authid/detail.uri?authorId=55625974200
- Web of Science: https://www.webofscience.com/wos/author/record/R-8828-2018