Dear Topaz team,
first of all, many thanks for developing and maintaining such a powerful tool!
I am facing an issue with recent versions of Topaz (0.3.7 and 0.3.17). The motion-corrected micrographs are preprocessed using Cryosparc wrapper that runs the following command:
topaz preprocess --scale 9 --niters 200 --num-workers 4 -o /.../preprocessed [12 MICROGRAPH PATHS EXCLUDED FOR LEGIBILITY]
Afterwards the picking fails with the error:
AssertionError: No particles extracted. Possible causes include 'Particle threshold' parameter being too high or a version discrepancy between the Topaz or PyTorch instance that trained the model and the one that is being used for extraction.
If I check the *.txt file produced by the wrapper, then it contains picks, but all picks have score nan. Hence none of them pass the threshold.
The issue becomes obvious when I open the preprocessed micrographs:
These same micrographs work without any problems in Cryosparc pickers.
If I apply Cryosparc denoiser on the micrographs, then Topaz preprocess works correctly:
Also, if I use an earlier version of Topaz (0.2.5a), preprocess works OK:
Lastly, if I use motion-corrected micrographs from EMPIAR-10025, Topaz 0.3.7 preprocesses & picks without any issues.
What could be the problem here? Perhaps, it has to do with the modifications of the preprocessing between Topaz versions 0.2.x and 0.3.x?
Dear Topaz team,
first of all, many thanks for developing and maintaining such a powerful tool!
I am facing an issue with recent versions of Topaz (0.3.7 and 0.3.17). The motion-corrected micrographs are preprocessed using Cryosparc wrapper that runs the following command:
topaz preprocess --scale 9 --niters 200 --num-workers 4 -o /.../preprocessed [12 MICROGRAPH PATHS EXCLUDED FOR LEGIBILITY]Afterwards the picking fails with the error:
AssertionError: No particles extracted. Possible causes include 'Particle threshold' parameter being too high or a version discrepancy between the Topaz or PyTorch instance that trained the model and the one that is being used for extraction.If I check the *.txt file produced by the wrapper, then it contains picks, but all picks have score
nan. Hence none of them pass the threshold.The issue becomes obvious when I open the preprocessed micrographs:
These same micrographs work without any problems in Cryosparc pickers.
If I apply Cryosparc denoiser on the micrographs, then Topaz preprocess works correctly:
Also, if I use an earlier version of Topaz (0.2.5a), preprocess works OK:
Lastly, if I use motion-corrected micrographs from EMPIAR-10025, Topaz 0.3.7 preprocesses & picks without any issues.
What could be the problem here? Perhaps, it has to do with the modifications of the preprocessing between Topaz versions 0.2.x and 0.3.x?