A processing plugin for using machine learning in QGIS.
It exposes the power of scikit-learn random forest through QGIS processing.
Random Forest is a robust, well-known machine learning algorithm for classification and regression tasks. It works by creating multiple decision trees during the training. The output is generated via majority voting in case of classification or the average of the prediction of the trees in case of regression. With this plugin, you can execute a Random Forest classifier that uses 300 estimators (trees). A random_state is set to seven for reproducibility.
The algorithm requires the following parameters:
- A point vector layer as input data for training
- The classification field that contains the class labels
- The image to process
- (optional) A name for the output - classified image
You only need to prepare your training samples with points and the class labels. The plugin will directly sample the raster to prepare the training data, which will be randomly split into training and testing sets.
Watch the video below on how to run ml4qgis
What's new?
You can now specify the attribute name for your labels or classes, even text attributes! ML4QGIS will take care of all necessary conversions.
The plugin samples your raster, splits the data into training and testing sets, and automatically displays the classified image in random colors.
Watch the video below for the Caracas release
This plugin is developed in collaboration with Rosa Aguilar (ITC/University of Twente) and Matthias Kuhn (OPENGIS.ch).