$"\"$ DGM lib

Machine Learning
with
Conditional Random Fields

Introduction

DGM is a cross-platform C++ library implementing various tasks in probabilistic graphical models with pairwise or complete (dense) dependencies. The library aims to be used for the Markov and Conditional Random Fields (MRF / CRF), Markov Chains, Bayesian and Neural Networks, etc. Specifically, it includes a variety of methods for the following tasks:

Learning: Training of unary and pairwise potentials
Inference / Decoding: Computing the conditional probabilities and the most likely configuration
Parameter Estimation: Computing maximum likelihood (or MAP) estimates of the parameters
Evaluation / Visualization: Evaluation and visualization of the classification results
Data Analysis: Extraction, analysis and visualization of valuable knowledge from training data
Feature Extraction: Extraction of various descriptors from images, which are useful for classification

DGM is released under a BSD license and hence it is free for both academic and commercial use.

Quick Links

Documentation
Installation guide
Tutorials
Fork us on GitHub
User Q&A forum
Blog
Report a bug

Fast

Low overhead, DGM has only one external dependency: OpenCV. Optimized for high-efficient calculations and takes advantage of multi-core processing as well as GPU computing.

Batteries Included

Comes out of the box with everything you need to create your first machine learning application. A selection of demo projects may serve as the basis for your own application.

Cross-Platform

DGM is a cross-platform, dynamic-link library, meant to be used in Windows, Mac and Linux. Its C++17 code is compiled with Microsoft Visual Studio, Xcode and gcc.

Application Examples

Remote Sensing

The first example considers a challenging problem in remote sensing: the precise classification and reconstruction of urban areas from aerial images. Here we took airborne color-infra-red images with the corresponding digital surface models (terrain elevation maps), recieved with an aircraft-based visual sensors. The DGM library was used here for joint object detection (e.g. cars (red color)) and Earth surface classification (grass (green color), road (gray color), sidewalks (blue color), etc.)

Please visit the project Web-Page for more details.

Bio-Medical Imaging

The algorithms for accurate eyelids, iris and pupil- extraction are all currently being of a high demand as they allow for challenging technological developments as robot-assisted surgery, biometrics, gaze tracking, etc. The input image here is a photo of an eye. The second image shows the per-pixel classification result and the third image – the same result after applying the DGM-based Conditional Random Field. Here, the CRF-based classifier has been trained only on 22 images, segmented manually.

Four different colors represent eyelids (yellow), sclera (red), iris (blue) and pupil (gray)

Stereo

In this example we apply DGM-based Markov Random Field to the stereo correspondence problem. The disparity map between left and right images of a stereo pair is estimated in form of a class map, where every class represents a small range of disparity values. The following result was achieved by applying approximate tree reweighted inference algorithm and Potts smoothness term, thus, no training was needed.

Please refer to Demo Stereo for the source code and more details.

Microbiology

The labeling of Environmental Microorganisms (EM) which help decomposing pollutants, plays a fundamental role for establishing sustainable ecosystem. Here we propose an EM classification engine that can automatically analyze microscopic images using DGM-based Conditional Random Fields. Our CRF model localizes and classifies EMs by considering the spatial relations among local and global features.

Images from left to right: Input image of Keratella Quadrata; Extracted local features; and resulting segmentation. Please visit the project Web-Page for more details.

If you have applied the DGM library in your work and achieved some interesting results, please send them to me via sergey.kosov@project-10.de, so I could allocate them at this web-page. This will help us to spread and improve the library. We thank you in advance for your support.

DGM in Publication

Selected main publications:
- Labeling of Partially Occluded Regions via the Multi-Layer CRF
  in Multimedia and Tools Applications, 2019
- A High Performance Implementation of A Unified CRF Model for Trust Prediction
  in IEEE 20th International Conference on High Performance Computing and Communications, 2018
- Environmental Microorganism Classification Using Conditional Random Fields and Deep Convolutional Neural Networks
  in Pattern Recognition, 2018
See also more publications, where the DGM libray was used:
- Three-dimensional conditional random field for the dermal–epidermal junction segmentation
  in Journal of Medical Imaging, 2019
- Segmentation of the proximal femur in radial MR scans using a random forest classifier and deformable model registration
  in International Journal of Computer Assisted Radiology and Surgery, 2019
- Classification of the dermal-epidermal junction using in-vivo confocal microscopy
  in proc. of 14th International Symposium on Biomedical Imaging (ISBI 2017)
- A higher order Conditional Random Field model for simultaneous classification of land cover and land use
  in ISPRS Journal of Photogrammetry and Remote Sensing, 2017
- Unsupervised Change Detection in High Spatial Resolution Remote Sensing Images Based on A Conditional Random Field Model
  in European Journal of Remote Sensing, 2016
- Spatial-temporal conditional random fields crop classification from TerraSAR-X Images
  in ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2015
- Land Use Classification using CRFs for the Verification of Geospatial Databases
  in ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, 2014
- A Two-Layer Conditional Random Field Model for Simultaneous Classification of Land Cover and Land Use
  the Int. Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2014
- Sequential Gaussian Mixture Models for Two-Level Conditional Random Fields
  in proc. of the 35th German Conference on Pattern Recognition (GCPR 2013)
- A Two-Layer Conditional Random Field for the Classification of Partially Occluded Objects
  Cornell University Library, arXiv:1307.3043, 2013
- The Application of a Car Confidence Feature for the Classification of Cross-Roads Using CRFs
  in ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, 2013
- 3D Classification of Crossroads from Multiple Aerial Images Using Conditional Random Fields
  in IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS 2012)
- 3D Classification of Crossroads from Multiple Aerial Images Using Markov Random Fields
  in proc. of the 22nd ISPRS congress (ISPRS 2012)

To reference DGM in a publication, please include the library name and a link to this website [BibTeX]. You may also want to include the library version, since we currently update the software.