24
S. P. Donegan and M. A. Groeber
from materials characterization or simulation was typically accomplished using
bespoke solutions tailored to a particular problem type. However, with the advent of
relatively inexpensive computational infrastructure, availability of modern statistical
and machine learning algorithms, and the popularity of open-source development,
several tools for materials data analytics have gained traction within the industrial
and research communities.
Since much of the characterization data collected in materials research take
the form of n-D images, many analytics tools have been developed specifically
tailored for image processing. Avizo™ is a commercial software package that
provides image processing and analytics capabilities for materials images, including
segmentation, computation of feature statistics such as size and shape, and meshing
[23]. Similarly, the commercial GeoDict ® software provides solutions for computed
tomography processing, fiber analysis, and synthetic composite simulation [24]. For
3D EBSD data, ESPRIT QUBE commercially provides solutions for reconstruction
and alignment, misorientation segmentation, and texture analysis [25].
Several open-source tools provide more general capabilities than the commercial
products described above. The Materials Knowledge System in Python (PyMKS) is
an open-source Python framework intended to provide data science approaches for
solving various materials problems [26]. PyMKS has support for a variety of analytics tailored to materials, such as microstructure quantification using 2-point statistics
[27] and fitting surrogate convolutional kernels to FEM data, producing highly
accelerated elastic models [28]. A similar toolset is the Materials-Agnostic Platform
for Informatics and Exploration (Magpie), an open-source Java-based library for
fitting various machine learning models to materials data [29]. Specifically for
texture analysis, the MATLAB toolbox MTEX provides capabilities for plotting
pole figures, segmenting grains, and computing orientation distribution functions
[30, 31]. Also leveraging MATLAB, the Materials Image Processing and Automated
Reconstruction (MIPAR™) software provides proprietary routines customized for
2D and 3D materials image analysis [32].
3.2 Example Tools from Other Fields
Materials is not the only field that must contend with multiscale, multimodal,
hierarchical information. Specifically, the medical and biomedical communities
often handle multimodal information streams, however with a focus on n-D images.
One of the most widely used tools for scientific biomedical image analysis is ImageJ
[34, 35]. Publically funded by the National Institutes of Health, ImageJ is a Javabased library and application that contains a wide variety of common and advanced
image processing methods. Fiji is a popular open-source distribution of ImageJ that
contains several additional plugins for advanced image analysis and segmentation
[36, 37]. Another popular library for medical image analysis is the open-source
Insight Segmentation and Registration Toolkit (ITK) [38, 39]. ITK, by taking
advantage of generic template programming techniques in C++, provides highly
S. P. Donegan and M. A. Groeber
from materials characterization or simulation was typically accomplished using
bespoke solutions tailored to a particular problem type. However, with the advent of
relatively inexpensive computational infrastructure, availability of modern statistical
and machine learning algorithms, and the popularity of open-source development,
several tools for materials data analytics have gained traction within the industrial
and research communities.
Since much of the characterization data collected in materials research take
the form of n-D images, many analytics tools have been developed specifically
tailored for image processing. Avizo™ is a commercial software package that
provides image processing and analytics capabilities for materials images, including
segmentation, computation of feature statistics such as size and shape, and meshing
[23]. Similarly, the commercial GeoDict ® software provides solutions for computed
tomography processing, fiber analysis, and synthetic composite simulation [24]. For
3D EBSD data, ESPRIT QUBE commercially provides solutions for reconstruction
and alignment, misorientation segmentation, and texture analysis [25].
Several open-source tools provide more general capabilities than the commercial
products described above. The Materials Knowledge System in Python (PyMKS) is
an open-source Python framework intended to provide data science approaches for
solving various materials problems [26]. PyMKS has support for a variety of analytics tailored to materials, such as microstructure quantification using 2-point statistics
[27] and fitting surrogate convolutional kernels to FEM data, producing highly
accelerated elastic models [28]. A similar toolset is the Materials-Agnostic Platform
for Informatics and Exploration (Magpie), an open-source Java-based library for
fitting various machine learning models to materials data [29]. Specifically for
texture analysis, the MATLAB toolbox MTEX provides capabilities for plotting
pole figures, segmenting grains, and computing orientation distribution functions
[30, 31]. Also leveraging MATLAB, the Materials Image Processing and Automated
Reconstruction (MIPAR™) software provides proprietary routines customized for
2D and 3D materials image analysis [32].
3.2 Example Tools from Other Fields
Materials is not the only field that must contend with multiscale, multimodal,
hierarchical information. Specifically, the medical and biomedical communities
often handle multimodal information streams, however with a focus on n-D images.
One of the most widely used tools for scientific biomedical image analysis is ImageJ
[34, 35]. Publically funded by the National Institutes of Health, ImageJ is a Javabased library and application that contains a wide variety of common and advanced
image processing methods. Fiji is a popular open-source distribution of ImageJ that
contains several additional plugins for advanced image analysis and segmentation
[36, 37]. Another popular library for medical image analysis is the open-source
Insight Segmentation and Registration Toolkit (ITK) [38, 39]. ITK, by taking
advantage of generic template programming techniques in C++, provides highly