Classifier of superpixels / supervoxels

Classifier of superpixels/supervoxels is a good method for automatic segmentation of images using train and predict scheme.

This classifier uses the SLIC (Simple Linear Iterative Clustering) algorithm written by Radhakrishna Achanta, Appu Shaji, Kevin Smith, Aurelien Lucchi, Pascal Fua, and Sabine Süsstrunk, Ecole Polytechnique Federale de Lausanne (EPFL), Switzerland to simplify the dataset by clustering of pixels into groups: superpixels for 2D or supervoxels for 3D. Each of these superpixels/voxels is characterized and these characteristics are used for the classification.

The example of use is presented below.

Back to Index --> User Guide --> Menu --> Tools


Dataset and the aim of the segmentation

Below is a dataset imaged with light microscopy where the aim is to segment the outlines of the cells (in green). The cells have different intensity and can't be directly segemented using the black-and-white thresholding.

Training the classifier

The first part of the classification is to select areas that belong to the object of interest and background.

In this window it is possible either load classifier from the training session done earlier (the Load classifier button), or train a new one if labels exist.

Wiping the temp directory

During the prediction the classifier creates files in the RF_Temp directory. This directory can be deleted by pressing the Wipe Temp dir button or manually using any file explorer.

Back to Index --> User Guide --> Menu --> Tools