A Robust and User-friendly Machine Learning Image Segmentation - - PowerPoint PPT Presentation

▶
a robust and user friendly machine learning image
SMART_READER_LITE
LIVE PREVIEW

A Robust and User-friendly Machine Learning Image Segmentation - - PowerPoint PPT Presentation

Nov 18, 2022 654 likes •1.08k views

Segmentation Trainer A Robust and User-friendly Machine Learning Image Segmentation Solution Presented by Mike Marsh, Ph.D. Dragonfly Product Manager Thursday, March 2, 2017 10th FIB-SEM Users Group Meeting Gaithersburg, MD About ORS

slide-1
SLIDE 1

Segmentation Trainer

A Robust and User-friendly Machine Learning Image Segmentation Solution

Presented by Mike Marsh, Ph.D. Dragonfly Product Manager Thursday, March 2, 2017 10th FIB-SEM Users’ Group Meeting Gaithersburg, MD

slide-2
SLIDE 2

About ORS

▸ Headquartered in Montreal, Canada. ▸ Founded in 2004. ▸ Registered users in 80 countries. ▸ Practicing ISO and IEC standards compliant processes

Visual SI ORS Visual

slide-3
SLIDE 3

About ORS

▸ Headquartered in Montreal, Canada. ▸ Founded in 2004. ▸ Registered users in 80 countries. ▸ Practicing ISO and IEC standards compliant processes

Dragonfly ORS Visual

slide-4
SLIDE 4

Rapid Innovation

V2.0 launched September 2016 V2.1 coming in April 2017 ▸ macro engine ▸ superpixel segmentation ▸ machine learning segmentation engine ▸ In-application store V2.2 coming in fall 2017 (coming to Linux)

Technology

Anaconda Python 3.5 for scientific computing State-of-the-art image segmentation High-impact rendering engine

Extensibility and Community

Sockets for extensions: ▸ Embedded online console ▸ Object analysis measurements ▸ Image filters ▸ Menu-actions ▸ Macros ▸ Machine Learning classifiers ▸ (and more) App store for sharing and versioning

Licensing

Flexible licensing options for various institutional needs Free licensing for non-commercial use in most countries

About Dragonfly

slide-5
SLIDE 5

Image Segmentation The hard way and the easy way

Painstaking: ▸ Painting ▸ Constrained Painting ▹ Threshold-gated painting ▹ Superpixel-bloc painting Easy, but never good enough: ▸ Point-and-click ▸ Thresholding (interactive) ▸ Thresholding (algorithmically, eg. Otsu’s method) ▸ Other tools ▸ Automated?

slide-6
SLIDE 6

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

Input image Class 1 Class 2 Segmentation

Classifier

slide-7
SLIDE 7

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

FIB-SEM of fuel cell Segmentation Electrolyte Electrode Pore space

Classifier

slide-8
SLIDE 8

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

FIB-SEM of fuel cell Segmentation Electrolyte Electrode Pore space

slide-9
SLIDE 9

Filter Banks

Use any of the filters in the Image Processing toolbox ▸ Smoothing ▸ Edge Enhancement ▸ Texture ▹ Gabor ▹ HoG ▹ DoG ▹ Standard deviation Aggregate into filter banks

slide-10
SLIDE 10

Spatial Discretization

slide-11
SLIDE 11

Spatial Discretization

slide-12
SLIDE 12

Spatial Discretization

slide-13
SLIDE 13

Pixel classification SmartGrid cell classification: ▸ Superpixel ▸ Watershed on Grid ▸ Superixel (Scikit-learn) ▸ Watershed on Grid (Scikit-learn)

Spatial Discretization

slide-14
SLIDE 14

Machine Learning Core

▸ Random Forest ▸ Extra-Trees ▸ Adaboost ▸ Gradient Boosting ▸ Bagging ▸ K-Nearest Neighbors

slide-15
SLIDE 15

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

FIB-SEM of fuel cell Segmentation Electrolyte Electrode Pore space

slide-16
SLIDE 16

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

Electrolyte Electrode Segmentation Pore space SE BSE

slide-17
SLIDE 17

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

Black box classifier segmentation

Electrolyte Electrode Segmentation Pore space SE BSE

slide-18
SLIDE 18

Machine learning core Engine Parameters Spatial discretization settings Smart Grid (Region) engine Parameters Filter bank (Feature Presets)

Train it

Classifier

Electrolyte Electrode Pore space Electrolyte Electrode Segmentation Pore space SE image BSE image

slide-19
SLIDE 19

Machine learning core Engine Parameters Spatial discretization settings Smart Grid (Region) engine Parameters Filter bank (Feature Presets)

Apply it

Classifier

Electrolyte Electrode Segmentation Pore space SE image BSE image

slide-20
SLIDE 20

Machine learning core Engine Parameters Spatial discretization settings Smart Grid (Region) engine Parameters Filter bank (Feature Presets)

Apply it

Classifier

Mask Electrolyte Electrode Segmentation Pore space SE image BSE image

slide-21
SLIDE 21
slide-22
SLIDE 22
slide-23
SLIDE 23
slide-24
SLIDE 24
slide-25
SLIDE 25
slide-26
SLIDE 26
slide-27
SLIDE 27
slide-28
SLIDE 28
slide-29
SLIDE 29
slide-30
SLIDE 30
slide-31
SLIDE 31
slide-32
SLIDE 32
slide-33
SLIDE 33
slide-34
SLIDE 34

Classifier

Machine learning core

  • Engine
  • Parameters

Spatial Discretization

Smart Grid (Region)

  • Engine
  • Parameters

Signal Textures

Filter bank (Feature Presets)

It’s modular

Electrolyte Electrode Segmentation Pore space SE image BSE image

slide-35
SLIDE 35

Classifier

Deep Learning core

  • Engine
  • Parameters

Spatial Discretization

Not necessary

Signal Textures

Not necessary

It’s modular (Deep Learning CNN) Late 2017

Electrolyte Electrode Segmentation Pore space SE image BSE image

slide-36
SLIDE 36

Encourage re-use of Classifiers

▸ Share classifiers with the community in the App Store (Infinite Toolbox) April 2017 ▸ Preview classifiers online Late 2017

slide-37
SLIDE 37

Acknowledgments

▸ Isabelle Bouchard ▸ Nicolas Piche ▸ scikit-learn.org (Machine Learning in Python)

slide-38
SLIDE 38
slide-39
SLIDE 39

Workflow for Using Classifiers

Build the classifier Train it Tune it Re-use it

slide-40
SLIDE 40

Workflow for Using Classifiers

Build the classifier

Train it Tune it ▸ Iterate: ▹ Update training classes ▹ Tweak engine parameters ▹ Add / remove filter banks review coefficients ▹ Retrain ▸ Preview Re-use it

slide-41
SLIDE 41

Segmenting Systematically (and with multiple signals)

1D thresholding: Use range 2D thresholding: Histographic segmentation 3D, 4D, ... : ??? ▸ BSE, ESB ▸ Elemental maps: Cu, Mb, Sn, Ni, ▸ More common than that: beyond simple signal intensity, you may have spatially correlated signal (e.g. texture)