mining site recognition

Recognize mining sites from satellite data. Download images of satellite data from given coordinates.

Structure

• classification/
  • classification.py: Train network on image data
  • global_parameters.py: Contains hyperparameters etc. to make it easier to test settings on different datasets
  • load_classification.py: Load all saved model data and plot results
  • save_augemented_images: To enable the writing of augmented images for inspection
• gid_database/
  • assign_GID_database.py: Rename tifs image files and assign global identifier
  • restructure_GID_database.py: Restructure the database and dissect the global identifiers into columns
• satellite_data_scraping/
  • image_config.py: Settings for images
  • sentinelhub_image_download.py: Download images via sentinelhub
  • sh_config_username.py: Example configuration file
• utils/
  • bulk_conversion.py: Conversion of satelite data from .tif to .png
  • image_splitting.py: Split data sets into train/valid/test

Usage

Satellite image download

1. Set up account with https://www.sentinel-hub.com/ to be able to download images
2. Get instance_id here
3. Get client_id and client_secret when setting up an OAuth client here
4. Save sentinelhub config file sh_config.py in satellite_data_scraping/.

   from sentinelhub import SHConfig
   
   config = SHConfig()
   
   config.instance_id = '{instance_id_here}'
   config.sh_client_id = '{sh_client_id_here}'
   config.sh_client_secret = '{sh_client_secret_here}'
   
   config.save()
   

   More infos here: Link
5. Setup dimensions, coordinates, bands etc. of image locations in image_config_{username}.py. Do not modify image_config.py!
6. Run sentinelhub_image_download.py

classification

1. Set-up folder /{project_name}/ (e.g. /infrared_images/) for classification of images with given settings
2. Save all images in folder /{project_name}/original_data/
3. Execute image_splitting.py in /{project_name}/ to sort images into train, validation and test folders
4. Copy all files from folder classification/ into **/{project_name}/
5. Adjust settings in global_parameters.py (now in /{project_name}/) if needed
6. Train model by running classification.py
7. Re-load model and view results with load_classification.py