Update: Semi-Automatic Classification Plugin v. 2.1.0

I have released a major update for the Semi-Automatic Classification Plugin for QGIS 2.0, version 2.1.0.
I have added several new features to the plugin and in particular to the interface: the ROI and classification commands are placed inside two movable docks, which allow for a better integration with the QGIS interface. This way, the map is better displayed for the ROI creation and the classification.

The Semi-Automatic Classification Plugin v. 2.1.0

The following is the main changelog:
Major update:
-changes to the interface, now ROI creation is a dock widget
 -new option for fast ROI on one band
-new tab for multiple ROI creation from point coordinates
-new Post processing tab that allows for:
-Accuracy assessment using a vector reference ground truth (it requires GRASS)
-Land cover change calculation between two classifications (it requires GDAL)
-new option to select ROIs for classification and preview
-new option to mask a classification using a shapefile  in classification tab
-new option to create a vector classification in classification tab
-added field names customization in settings tab
-added test functions in settings tab
-version for QGIS 2.0

Input files and ROI parameters are placed in the new dock ROI creation. A new option allows for the creation of ROIs using only one image band, expediting the ROI computation (that otherwise is computed using all the image bands) which is ideal for hyperspectral images.
ROI creation dock

The new dock Classification displays the ROI list, allowing for the selection of a ROI subset used for the preview and the classification processes. This way, it is possible to assess rapidly the effect of every ROI in the classification results. Also, this list allows for the management (delete selected ROIs) and visualization of ROIs (double click an item to zoom to the ROI).
Other new features are: the ability to create the output classification as a shapefile; the ability to calculate the accuracy (see below Post processing); and the ability to mask the classification output using a shapefile mask.
Classification dock

Several other tools are contained in the main plugin interface, such as the Spectral signature (this is not a new feature), and the new functionality Multiple ROI creation, which allows for the automatic creation of ROIs, given a list of points (coordinates) and class definitions. This can be useful for the rapid classification of multi-temporal images, or for creating several ROIs using GPS data from field survey.
Spectral signature tab

Multiple ROI creation tab

The Post processing tab provides the new abilities to perform the Accuracy assessment, comparing a classification to a reference shapefile (using the GRASS tool r.kappa), and to calculate the Land cover change between two classifications (using GDAL and Numpy).
Accuracy tab

Land cover change tab

Now, it is possible to customize the names of the training shapefile's required fields, in the Settings tab. This way, one can use any existing shapefile containing training polygons.
Other new functions are about debugging, in particular for testing the installation of the several dependencies (SAGA GIS, Orfeo Toolbox, GRASS GIS).
Settings tab

Soon I will post a video tutorial about these new features.
You can download the updated version of the Semi-Automatic Classification Plugin from here, and it will be soon available through the QGIS repository.
Please note that QGIS 2.0 is still under development, therefore some changes could affect the plugin until the final version is released.