Tutorials

How to use the code

The documentation for DGM consists of a series of demos, showing how to use DGM to perform various tasks. These demos also contain some tutorial material on graphical models.

Bacis Tutorials

• Demo 1D : An introduction to graphical models and to the tasks of inference and decoding on a set of simple examples:
  • Exact : An introduction to graphical models and the tasks of decoding and inference on a small graphical model where we can do everything by hand.
  • Chain : An introduction to Markov independence properties on an example of a chain-structured graphical model, and to efficient dynamic programming algorithms for inference.
  • Tree : This demo shows how to construct a tree-structured graphical model, for which also an exact message-passing inference algorithm exists.
• Demo 2D : An example of more complicated graphical models, containing loops and built upon a binary 2-dimentional image. This example also shows the application of DGM to unsupervised segmentation.
• Demo Dense : An introduction to the complete (dense) graphical models. The application of regular edge potentials used for pairwise graphs makes the inference practically impossible, thus special edge models for dense graphs should be applied.
• Demo Stereo : An example of CRFs application to the problem of disparity estimation between a pair of stereo images.
• Demo Feature Extraction : An introduction to the feature extraction, needed mainly for supervised learning.
• Demo Train : An introdiction to the random model learning (training) in case when the training data is available.
• Demo Visualization : An example of usage the visualization module of the library for analysis and represention of the intermediate and final results of classification.

Advanced Tutorials

• Demo Random Model : An advanced tutorial to the unary potentials training.
• Demo Parameters Estimation : An example of applying the model parameters estimation to a real problem.