Home

Documentation is for version 0.71.

ASEMI Segmenter is a program for automatically segmenting different materials in a 3D tomograph such as bones, textiles, air, and so on. It works by learning what each different material looks like from a few manually segmented slices in the volume, followed by automatically applying what was learned from the manually segmented slices on the rest of the volume.

This wiki provides information about the program together with user guides. The program can be either used as a stand alone command line program or as a Python library.

User guides

For a general introduction to segmentation and how this program works you can see the general information page. For instructions on how to install the program you can see the installation instructions page.

The following links take you to explanations of the different commands and input formats of the program:

• Preprocess command: Preprocess the tomograph volume into a format that can be used by the other commands.
  • Volume format: What this command expects to see as a volume.
  • Preprocess configuration: How to configure the preprocess command.
• Analyse data command: Get information about a data set.
  • Data analysis result format: What results consist of.
  • Data analysis configuration: How to configure the data analysis command.
• Tune command: Search for a train configuration that works on a particular dataset.
  • Tune configuration: How to configure the tune command.
• Train command: Train the segmenter using examples of manually labelled slices from the volume.
  • Dataset format: What this command expects to see in manually labelled slices.
  • Train configuration: How to configure the train command.
• Evaluate command: Evaluate the trained segmenter using manually labelled slices as reference.
  • Evaluation result format: What the results consist of.
• Segment command: Segment a volume using a trained segmentation model.
  • Segment configuration: How to configure the segment command.
  • Segmented volume format: What the resulting segmentation consists of.

The following links conern all commands:

• Log files: How commands keep a log of their activity in a text file.

Technical documentation

• Checkpoint files: How commands can continue from where they left off in case of interruption.
• Preprocessed volume format: What the resulting output file consists of.
• Training set file format: The intermediate file containing information for training the segmenter model.
• Trained model format: What the resulting output model file consists of.
• Features table file format: What the feature cache table consists of.

About the program

The Automated SEgmentation of Microtomography Imaging (ASEMI) project is a collaborative research project between the University of Malta and the European Synchrotron Radiation Facility (ESRF). The project has received funding from the ATTRACT project funded by the EC under Grant Agreement 777222.