ilastik Import Export plugin in FIJI

This plugin allows you to save images opened in FIJI directly to the hdf5 format preferred by ilastik. While this conversion is beneficial for any large dataset, we especially recommend it for multipage tiffs. After you save the file and process it with ilastik, you can use the same plugin to load the results for further post-processing in FIJI.


In FIJI, go to Help->Update… In the dialog that appeares, click the “Manage update sites” button. In the list of update sites, select “ilastik Import Export”. You should now see ilastik_HDF5.jar in your updater window.

From FIJI to ilastik

This is how you find the right export option from within FIJI. It will warn you to be careful with the axes order, but don't worry, if you get it wrong you can still correct it on import in ilastik.


Open the file as usual. If the axis order is wrong, as shown in this figure (tzyxc instead of txyzc), change it in this dialogue.

Train the classifier in ilastik as you usually would (see Pixel Classification workflow docs if you are not sure how to proceed). Once you are happy with the results, switch to the export applet.

Depending on your post-processing, export Probabilities or Simple Segmentation. You can control the export source by the “Source” dropdown menu. For probability maps you can either leave all settings at default, or, if you need to save space, convert the results to unsigned int 8-bit and rescale them from [0.0, 1.0] interval to [0, 255].

From ilastik to FIJI

Using the same plugin, you can now load your results into FIJI by selecting "ilastik HDF5" in the File->Import menu. Two options are available: load (and display) the raw pixel values or load and display the image with a lookup-table (LUT).

For Probabilities, choose Load Raw.
For Segmentation, choose Load LUT.
If you loaded the segmentation and the image looks black, you probably loaded it as raw. Try again with a LUT.

If you load Probabilities, you should see something like the left image. For Simple Segmentation with a lookup-table you should see the right image.