Loading data from a single file

You can use the “Add separate image(s)…” option to add new data files to a project. This will present a standard file open dialog, where the desired input file can be selected. This dialog also allows selecting multiple files. Note that multiple files will be added as if the dialog was used once for each of these. See the following section for instructions on how to interpret multiple files as a new dimension in the data.

Loading image stacks

If one dimension of the data is separated into different files, such as a stack of 2D images to form a 3D image or a sequence of 2D images for each frame in a movie, this option can be used to form a new dimension based on a list of files.

Selecting this option will open a new dialog to select the files that will be included in the stack. The selection can be specified in three ways:

• selecting multiple files using the standard system file open dialog,
• selecting all files in a directory
• using a filename pattern (with Unix style patterns)

A special case are image stacks saved in multiple .h5 files (HDF5, see paragraph Supported File Formats). HDF5 supports saving multiple datasets encapsulated in a single file which can be accessed using internal paths, similar to paths in a file system. For instance, you could have a folder with multiple files, file_001.h5, file_002.h5, ..., containing data that can be composed into a single image volume. In order to assemble the images an internal path to the image data in each file has to be specified. Ilastik assumes that data with similar internal paths should be stacked and will try to determine a common internal path in the selected .h5 files. In case of multiple common internal paths the user is asked to select the appropriate one. So, if the image data is found in file_*.h5/data/imagedata in each of the files, ilastik will automatically choose data/imagedata. In case of multiple datasets per file, e.g. the files might contain data collected from multiple subjects at internal paths like file_*.h5/images/subject1, file_*.h5/images/subject2, ..., ilastik will ask which of those sub-images should be selected for stacking.

Once a selection has been made, the File List box can be used to review the names of the files that will be imported as an image stack.

Note: Data that is loaded from image stacks (multiple files) will be copied to the project file (.ilp). This ensures faster computations and enables processing of stacks larger than RAM by converting the data to chunked hdf5.

Loading multiscale data

Starting in version 1.4.1, multiscale datasets are supported in the formats OME-Zarr (URLs containing .zarr) and Neuroglancer Precomputed (URLs starting with precomputed://).

You can use “Add multiscale dataset” to load multiscale, a.k.a. pyramidal images. The dialog will ask for an address. This must be a full URL including protocol. When you click “Check”, ilastik will try to obtain image metadata from the given address, and display the results of the request. If the dataset is stored on the local filesystem, you can paste the path into the address field, e.g. C:\Users\me\Downloads\tissue-gfp.zarr. The “Check” button will test whether the path exists on the filesystem and automatically convert it to a file:/// URL if successful. Note that this has to be the path to the root of the dataset, i.e. the folder containing the .zattrs file, and the folder name has to contain “.zarr”.

Note: ilastik will freeze while waiting for a response. If the server is slow or the connection is bad, it may take a while and your computer might warn that ilastik is not responding.

If the check is successful, you can confirm by clicking “Add to project”. You can then select which scale / resolution level to load using the drop-down box in the input data table.

To inspect more than one scale at a time, you can add multiple instances of the same dataset using “Add New”. Simply paste the same address again, click “Check”, and then “Add to project”.

When navigating the dataset, remember that downloading the image data can take some time. It is recommended to use the lowest resolution possible for exploration. At larger resolutions, zoom in to the area of interest, as suggested in the performance tips section. This way, ilastik can avoid loading data outside the viewer area.

Once you continue to the next step in the workflow, your choice of scale will be locked for all multiscale entries in the table. This is because ilastik can currently not handle changing pixel dimensions once a dataset is loaded into the workflow beyond “Input Data”. If you want to inspect a different scale later, you can use “Add new” and paste the same address again.