Prerequisites¶
A. System requirements¶
We currently support Windows, MacOS, and Linux operating systems. The minimum system requirements are:
8GB of RAM
8 CPU Cores
System storage: at least the same amount of free space as the size of your training dataset
1 NVIDIA GPU with 8GB of VRAM (optional)
Tip
A GPU is highly recommended for training models and make segmentation predictions.
Note
If you plan to use the plugin without a GPU, the training will default to using your CPU and will be significantly slower
Additionally, factors such as image size, 2D vs. 3D, resolution, model size, etc. may cause prediction to run slowly without a GPU
B. Data requirements¶
To use the plugin effectively, you should have the training data as following:
Existing high-quality segmentations, along with their corresponding raw original microscopy images
Large number of training images are recommended for robust, accurate segmentation models
This number depends on cellular structure type and their counts wihtin a single FoV or file
- The Segmenter ML plugin supports:
Image data format:
.tiff,.ome-tiffDimensions: 2D or 3D, single- (recommended for faster processing) or multi-channel
Segmentation object: cellular structures (recommended)
Tip
We recommend using the Allen Cell & Structure Segmenter Classic plugin to generate segmentations, as it ensures high-quality segmentations and compatibility with the required directory structure for this plugin.
If you would like to check out example data that works with the plugin, we share the data we have used for the pre-trained models provided with this plugin. Please refer to the Pre-trained models page for more details.