In digital image processing and computer vision, image segmentation refers to the process of partitioning a digital image into multiple segments or related sets of pixels.  This tutorial will introduce you to some basic image segmentation techniques driven by unsupervised machine learning techniques such as the Gaussian mixture model and k-means clustering. You will learn how to implement k-means clustering and template matching, and use these to segment a scanning electron microscopy image of graphene on a substrate. This tutorial also introduces you to U-Net, a popular convolutional neural network commonly developed for image segmentation in biomedicine. Using an assembled data set, you will learn how to create and train a U-Net neural network, and apply it to segment scanning electron microscopy images of graphene on a substrate.

Researchers should cite this work as follows:

• Aagam Rajeev Shah, Darren K Adams, Mitisha Surana, Ricardo Toro, Sameh H Tawfick, E. Ertekin (2021), "SEM Image Segmentation Workshop," https://nanohub.org/resources/imagesegment. (DOI: 10.21981/TFWE-DD75).

  BibTex | EndNote

1. convolutional neural networks (CNN)
2. Gaussian mixture model
3. graphene
4. image processing
5. image segmentation
6. k-means clustering
7. machine learning
8. materials science
9. nanoMFG Node
10. NCN Group - Machine Learning
11. NCN Group - Machine Learning Hands-on
12. scanning electron microscopy (SEM)