new workflow in beta: Trainable Domain Adaptation11 May 2023
For beta version
1.4.1b1, we have added a new workflow that combines shallow and deep learning approaches.
Check out the documentation for the Trainable Domain Adaptation Workflow.
In this workflow the Pixel Classification approach is combined with pre-trained convolutional neural networks (CNNs) for specific tasks. Currently, pre-trained networks are available for mitochondria segmentation in EM, and generic em boundary enhancement in EM, both for 2D and 3D data. The networks are hosted on BioImage Model Zoo, and are available automatically from within ilastik in this workflow.