satdetect : python object recognition from satellite images.
This Python package provides an open-source pipeline for training and evaluating object recognition systems for humanitarian purposes.
To train a detector
python -m satdetect.TrainDetector path/to/trainimage.jpg path/for/output/ [--OPTIONS]
To run a trained detector on new imagery
python -m satdetect.RunDetector path/to/image.jpg path/to/classifier.dump [--OPTIONS]
For more info, see TODO.
We require these python packages
- numpy
- scipy
- sci-kit learn
- sci-kit image
- joblib (for high-quality load/dump of Python objects to disk)
- satdetect/ : module code
- tests/ : unit-tests for module code
- kmlutil/ : utilities for converting between KML object annotations and flat files
- assets/ : template files for generating reports and KML visualizations