Image analysis

Materials characterization using deep learning image analysis

Sammy Nordqvist1* (S. Nordqvist), John Sosa2 (J. Sosa), Michael Kudlinski2 (M. Kudlinski), Alisa Stratulat2
(A. Stratulat), Dr. Simone Langner3 (S. Langner)
1 SciSpot, Aspvägen 13, Stenungsund, Sweden, 2 MIPAR Image Analysis Software, 8050 N High St, Ste 170,
Columbus, OH 43235, United States, 3ADDITIVE Soft- und Hardware für Technik und Wissenschaft GmbH, MaxPlanck-Str. 22b, D-61381 Friedrichsdorf, Germany *sammy@scispot.se

Background

For scientists and engineers across various disciplines, image analysis serves as a vital tool for gaining
deeper insights into complex phenomena. Whether studying biological structures, geological
formations, engineering components, or materials microstructures, the ability to extract quantitative
data from images is invaluable. Image analysis enables engineers and researchers to replace manual
analysis and increase productivity, gain additional data insights improve statistics, and advance
their work by solving complex problems. By harnessing the power of image analysis, scientists and
engineers can accelerate research, optimize manufacturing processes, and improve the performance of
products across a wide range of fields, ultimately advancing our understanding of the natural world and
improving technology.
Figure 1. Next-generation detection using real-time processing leads to improved detection visualization (right side).

Method and results

Using deep learning and powerful image analysis engines, MIPAR (www.mipar.us) allows users to
perform fast, accurate, and automated analysis of images. Through a user-friendly interface, engineers
and researchers can personalize analysis to their samples as well as visualize and extract measurements
– all without programming. Through five integrated applications, MIPAR offers flexibility and efficiency
for 2D and 3D analysis applications. The key ingredients are in the Recipes, which include a series of
analysis steps that are tailored to each application. As a result, researchers and engineers can now easily
solve problems such as grain size analysis excluding twins, particle analysis of clusters, defect
identification, and many more.
Seamlessly integrated within the MIPAR ecosystem, the Snap tool, powered by Spotlight, offers
accelerated creation of training datasets for an efficient and streamlined model training workflow. The
accurate segmentations provided by Spotlight allow for less time spent creating models and algorithms,
leading to faster data collection. MIPAR Spotlight simplifies image analysis further by limiting the
need for custom model configurations and amplifying a model’s ability.

Conclusion

This presentation will overview the advantages of using MIPAR Spotlight, the new cutting-edge of
detection, to analyze grains, particles, droplets, and defects with real industrial applications