UntangleWorms Training Erros

I downloaded the demonstration file for UntangleWorms. While the test data (2 images) work perfectly, the trained model does not give me the results I want. Therefore, I tried training for my own dataset. The diameter of the objects I am interested in are larger than the predefined sizes in the demonstration. I changed them to min = 40 and max = 100.
The output shows the worm nicely identified…
drive.google.com/folderview?id= … sp=sharing

Traceback (most recent call last):
File “cellprofiler\pipeline.pyc”, line 2307, in post_group
File “cellprofiler\modules\untangleworms.pyc”, line 727, in post_group
File “numpy\linalg\linalg.pyc”, line 445, in inv
File “numpy\linalg\linalg.pyc”, line 328, in solve
LinAlgError: Singular matrix

What am I doing wrong?


I think we need some more information to help, since I think this problem is unusual.

Would you post your pipeline plus an example image of yours so we can take a look?
Does the error happen with every image? After what step does it occur? Do you get this error with the example images or only yours?



Thank you very much for the reply.

I was able to find a work around to this. The problem was because only 1 worm was present in each image

I added an extra worm to each image (rotated or flipped, etc.) and the trainer works without a problem.

Is there a way to train datasets with just 1 worm each image?

Thank you

Additional Information:
Sample Image: dropbox.com/s/ksw3dhrl3qmuoo7/1.tiff
Pipeline for training: dropbox.com/s/1f0b2bst2tpp3 … ain.cppipe

The error seems to indicate that your training set was built using very few worms (or just a single worm?). I wasn’t able to download your pipeline, you can email it to me directly if you have difficulty posting (leek at broadinstitute dot org). Just to check, does your pipeline process all of the images during analysis or do you run the pipeline on each image individually?