interactive image segmentation with ilastik

Dominik Kutra, EMBL Heidelberg; Anna Kreshuk, EMBL Heidelberg
I2K 2022 | Workshops #1 (pre-sessions) | May 6th

ilastik is an easy-to-use interactive tool that brings machine-learning-based (bio)image analysis to end users without substantial computational expertise. It contains predefined workflows for image segmentation, object classification, counting and tracking.
In this tutorial, we will focus on the segmentation problem and take a close look at the Pixel Classification, Autocontext and Neural Network workflows. We will train multiple classifiers together by interactively labelling the images, apply them in batch mode. We will also explore how to find and apply pre-trained neural networks from the bioimage.io model zoo.

Pre-Workshop Instructions: Please download the lates beta-version of ilastik (1.4.0b27) for your platform: https://www.ilastik.org/download.html...
Also, please have a computer mouse ready.

Keywords: image analysis, segmentation, deep learning, machine learning

0:00 Talk: Introduction to ilastik, Pixel and Object Classification
22:12 Practical: Pixel Classification, creating a project, loading data, basic navigation
24:44 Practical: Pixel Classification, selecting pixel features
28:05 Practical: Pixel Classification, classifier training
47:45 Practical: Pixel Classification, exporting data
53:17 Practical: Object Classification
54:21 Practical: Loading ilastik hdf5/h5 files in Fiji
55:41 Practical: Object Classification, creating a project, loading data
56:40 Practical: Object Classification, thresholding, size filter
58:39 Practical: Object Classification, selecting object features
1:01:39 Practical: Object Classification, classifier training
1:05:22 Practical: Object Classification, exporting data
1:07:56 Talk: more ilastik workflows: Carving, Counting, Tracking, Deep Learning with Neural Network Classification, Boundary-Based Segmentation
1:25:40 Practical: Neural Network Classification Workflow
1:31:26 Practical: Boundary-based Segmentation with Multicut Workflow
1:40:22 Practical: Running ilastik Pixel Classification from Fiji