Failure to generate export file after tracking

I am receiving the following error message when trying to export tracking results:

Failed to generate export file:
file location and name
dtype uint32 does not have a matching OME-XML type

I have imported a number of pictures in a tiff format into a tracking project (with pixel prediction map) along with pixel classification export from a previous project in multipage tiff format. I went through the tracking steps with no problem but receive the above error message when trying to export in a multipage tiff format.

Any help would be very appreciated!

Hi @will.singleton,

welcome to the forum!

This error is sort of puzzling to me and of course we’d like to get to the root of it.
In the meantime, as a quick solution, you might consider exporting to a different format - we prefer hdf5 - and importing/converting it in fiji. The hdf5 file can be loaded into fiji using our fiji import/export plugin.

For the original issue; I tried reproducing it by exporting the tracking result from a 2D time series into multi-page tif, but it didn’t throw this error. Would you be able to provide a little more detail, like

  • what ilastik version did you use
  • are you on linux, macosx, or windows?
  • are you working with a 2D, or 3D time series?
  • what output are you exporting (export source)?


Hi Dominik,

Thank you for the quick response. Unfortunately, I am following a protocol so I am not sure if converting to hdf5 will be possible. It is probably worth mentioning that others have followed the steps I have taken without this error.

I am using version 1.3.3post3 on windows.
I am working with a 2D time series.

Source Image Description:
Shape: (57,2,700,1000) Axis Order: tcyx Data Type: uint32

Output Image Description: the same as source

Format: multipage tiff

Many thanks,

Hi @will.singleton,

thank you for providing the additional detail, I hope I can reproduce it now, maybe one more detail - how (with witch software) did you create those images?

Saving to hdf5 will not change the data (all values are preserved) but it will write it to the hard-drive slightly differently, such that it is easier to read in blocks (rather than lines).

Hi Domink,

Thank you for the hdf5 advice.

The images were taken directly from a camera integrated into a microscope which was controlled by a custom labview code which I think were saved as raw images.

Many thanks,

Hi Dominik,

Just a quick update - exporting as hdf5 and converting in FIJI worked.

Thank you for your help

1 Like