Open ovalerio opened 1 month ago
k-dominik
commented
1 month ago Hi @ovalerio,
sorry that ran into issues with the ilastik output in MoBiE - it's so nice when it works...
Do you think you could share the images with us?
I hope we can resolve this quickly
Cheers Dominik
ovalerio
commented
1 month ago Hello Dominik:
Certainly. =)
In Github I am restricted to max 25 MB file upload size. Therefore, I uploaded the files here
For context, I am creating the chunk files using h5py:
import h5py
import tifffile as tff
data = tff.imread(image_sequence.files[substack['z'][0]:substack['z'][1]])
## ilastik performance optimization: convert data into h5 file ##
chunk_name = 'chunk' + str(i)
chunksize = substack['z'][1] - substack['z'][0]
ilinp_h5 = io.join(res_dir,chunk_name +'.h5')
with h5py.File(ilinp_h5, 'w') as f:
f.create_dataset('exported_data', data=data, dtype='uint16', chunks=(min(64, chunksize), 64, 64))Thanks! ~ Omar
k-dominik
commented
1 month ago Hello @ovalerio,
thank you so much for providing the data, and apologies for the delayed response!
Also, your details on how you created the h5 files in the first place was super helpful! You're very close, even more of a pity it took me so long to reply :/
The reason MoBiE cannot load the data is that it is missing some attribute that helps MoBiE understand the meaning of the axes in your data.
Your data has 3 dimensions, so I assume it should be likely either "tyx" or "zyx". From the exported files I can see that you treated your data as 3D (3 spatial dimensions) ilastik. Generating the compatible metadata would entails somehting like this:
import h5py
# optional, in case you don't mind adding this dependency
import vigra
# then open your h5 file
with h5py.File("chunk0.h5", "a") as f:
f["exported_data"].attrs["axistags"] = vigra.defaultAxistags("zyx").toJSON()
# if you don't want to use vigra, so also no need to import at the beginning,
# no problem, you would alternatively do:
axistags = """
{
"axes": [
{
"key": "z",
"typeFlags": 2,
"resolution": 0,
"description": ""
},
{
"key": "y",
"typeFlags": 2,
"resolution": 0,
"description": ""
},
{
"key": "x",
"typeFlags": 2,
"resolution": 0,
"description": ""
}
]
}
"""
with h5py.File("chunk0.h5", "a") as f:
f["exported_data"].attrs["axistags"] = axistags
# for the rest of the input files, you'd do the same. The exported probabilities have this metadata attached.also, reference to our jupyter notebook for h5 conversion: https://github.com/ilastik/ilastik/blob/main/notebooks/h5convert/convert_to_h5.ipynb
Hope this helps. And please if you have any further questions, let me know.
Cheers Dominik
ovalerio
commented
1 month ago Hello Dominik,
Thank you. That was the solution. I am now able to open the h5 images using MoBIE. :rocket: :+1:
I am getting an error in the log.
Before that error the log show messages about two missing files. (keymaps.yaml, appearance.yaml)
[Fri Oct 04 09:41:47 UTC 2024] [ERROR] [] Command errored: Open Multiple Images and Labels...
java.lang.NullPointerException
at org.embl.mobie.lib.bdv.ImageNameOverlay.addImage(ImageNameOverlay.java:168)
at org.embl.mobie.lib.bdv.ImageNameOverlay.updateImages(ImageNameOverlay.java:162)
at org.embl.mobie.lib.bdv.ImageNameOverlay.setActive(ImageNameOverlay.java:111)
at org.embl.mobie.lib.view.ViewManager.show(ViewManager.java:309)
at org.embl.mobie.MoBIE.initUiAndShowView(MoBIE.java:288)
at org.embl.mobie.MoBIE.openImageAndLabelGrids(MoBIE.java:231)
at org.embl.mobie.MoBIE.<init>(MoBIE.java:175)
at org.embl.mobie.command.open.OpenMultipleImagesAndLabelsCommand.run(OpenMultipleImagesAndLabelsCommand.java:103)
at org.scijava.command.CommandModule.run(CommandModule.java:196)
at org.scijava.module.ModuleRunner.run(ModuleRunner.java:165)
at org.scijava.module.ModuleRunner.call(ModuleRunner.java:125)
at org.scijava.module.ModuleRunner.call(ModuleRunner.java:64)
at org.scijava.thread.DefaultThreadService.lambda$wrap$2(DefaultThreadService.java:247)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:750)I wonder if it can be fixed because it seems to slowdown the MoBIE viewer. This is the version of ImageJ I am currently running:
(Fiji Is Just) ImageJ 2.14.0/1.54f; Java 1.8.0_322 [64-bit]; Linux 3.10.0-1160.el7.x86_64; 5728MB of 130953MB (4%)For completeness I was running ImageJ as a superuser (root). I needed that to perform the plugin updates.
Best, Omar
tischi
commented
1 month ago Hi @ovalerio,
Could you please share the files and how you exactly open them in MoBIE such that we can try reproducing this error?
Thanks!
Dear
MoBIEdevelopers,Thank you for creating this Plugin.
I learnt about it while researching about QC tools for Pixel Classification jobs in
ilastik.As you can see in the
FIJIlog, I tried to open my images withMoBIE:The problem is that I got an IllegalArgumentException:
I am using
ilastikin headless mode. Here is the log for one of the substacks:Thanks in advance for your help.