16 Jan 2025

Deploy ROI Extractor and Patch Extractor Modules

This Xtra allows for ROI Extractor and Patch Extractor modules deployment.

MitochondriaSeg

To minimize the effort required by a user to train a deep learning model, in the new Segmentation+ Workroom you can define sparse annotations using Regions of Interest (ROI sets). Thus, you can efficiently train a deep learning model to segment an image starting from minimal annotation. Based on the defined ROIs and annotations, you can use the "AI Assisted Segmentation” tool in the segmentation editor. This tool has predefined parameters that allow efficient training of a deep learning model without the user needing advanced knowledge in the field.

In this tutorial, we will use the ROI Extractor and Patch Extractor modules in the project room, which allow an advanced user to specify data augmentation parameters. Starting from sparse annotation, augmentation helps create new data that is used during the training phase. The resulting model will be more generic and performant.

Installing this Xtra deploys the ROI Extractor and Patch Extractor modules, which are not present by default in Amira-Avizo Software.

The image used in this Xtra project is publicly available [1].

[1] Mitochondria Detection in EM Stacks. cvlab.epfl.ch/data/data-em/