README.md 1.31 KB
Newer Older
Himanshu's avatar
Himanshu committed
1 2
# LandCoverClassification-Sentinel2

Himanshu's avatar
Himanshu committed
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
Random Forest classification over Sentinel-2 data using training labels from OSM landcover database.

Prerequisites:

    osm-lulc database dump and PostgreSQL (part of preprocessing script)
    atleast 50GB RAM (for current subset size, or use lower subset size)
    12 CPU cores (or change the values in script)
    GDAL command line tools version >= 2.1.3
    Python version 2.7 and associated libraries
    R and associated libraries
    
Code structure:

    --downloadScripts   : scripts for download of Sentinel-2 data
    --flowchart         : for an overview of the workflow
    --logs              : preprocessing and RF logs for tiles in EU
    --reports           : pdf reports for the project (only for version 1 of code)
    --scripts           : classification scripts and others

Status:

    Total tiles in EU                       : 865
    Classified                              : 843
    Not classified (NO scenes for download) : 22

    
Refer to the flowchart to understand the workflow.  
For the whole process, two scripts need to be run separately:

    downloadScripts/all_tiles_download.R    : to download tiles
    scripts/RandomForest_main.py            : to run RF on downloaded tiles

After downloading the tiles, they need to be moved to RF script's input data location.