empanada-napari modules#
Note
“Run Quantized Model” is only supported for devices running on Intel Chips.
Inference modules#
Runs model inference on 2D images. Supports batch mode for predicting segmentations on a series of unrelated images or can be used to segment arbitrary 2D slices from volumetric data.
Implements stack and ortho-plane inference functionality for volumetric datasets.
Exports 2D stack segmentations or a single 3D volume label mask.
Finetune and training modules#
Automatically picks patches of data to annotate from 2D or 3D images. Also gives the option for uses for manually select ROIs using placed points.
Stores training patch segmentations in the correct format expected for model finetuning and training.
Allows users to finetune any registered model on a specialized segmentation dataset.
Train models from scratch for arbitrary panoptic segmentation tasks. Optionally, initialize training from CEM pre-trained weights for faster convergence and greater robustness.
Make a new model accessible in all other training and inference modules. Models can be registered from .pth files or from web URLs. Useful for sharing models locally or over the internet.
Get information about registered models to help decide which one is appropriate for inference or finetuning.
Locally exports an empanada model. Useful for sharing models locally or over the internet.
Makes a new model accessible in all other training and inference modules.
Archives a model into a hidden directory removing it from models menu
Proofreading modules#
Allows the selection of multiple instances and merges them all to the same label.
Allows the selection of multiple instances and allows the removal of selected labels
Given a label ID, moves the napari viewer to the first 2D slice where an object appears.
Returns the next available label ID for manual annotation.
Applies morphological operations to specific labels or to the entire label layer.
Counts the number of label IDs in the dataset and export the list of label IDs in an Excel workbook.
Removes small pixel/voxel valued labels from the label mask and/or labels touching the border of the image.
Allows the placement of multiple markers for distance watershed-based instance splitting.
Chop big images (with/without corresponding masks) to square tiles of given size
Merge square tiles back created by Create Tiles module