Extract from micrographs after Motioncorr2 failed: "Micrograph shapes are inconsistent!"

Hi everyone,

I tried to compare Motioncorr2 vs Patch motion correction using following pipeline: import particles, Motion correction (Patch / Motioncorr2), Patch CTF, Template picking, Extract from micrographs & Ab initio resonstriction & refinement.

However, the Motioncorr2 path has failed on “Extract from micrographs” with following log:

[CPU: 620.7 MB]  Traceback (most recent call last):
  File "cryosparc_worker/cryosparc_compute/run.py", line 84, in cryosparc_compute.run.main
  File "/opt/cryosparc/cryosparc_worker/cryosparc_compute/jobs/extract/run.py", line 153, in run_extract_micrographs_cpu_parallel
  File "/opt/cryosparc/cryosparc_worker/cryosparc_compute/jobs/extract/extraction_cpu_parallel.py", line 49, in do_extract_particles_single_mic_cpu
    assert mic_shape == mic.shape, "Micrograph shapes are inconsistent!" # otherwise micrograph_blob/shape is inconsistent with the actual frame shape
AssertionError: Micrograph shapes are inconsistent!

While the CryoSPARC default path was fine.

The Import movies input parameters are:

[CPU: 242.0 MB]  Loaded 977 movies.
[CPU: 242.0 MB]    Common fields: 
[CPU: 242.0 MB]                 mscope_params/accel_kv :  {200.0}
[CPU: 242.0 MB]                    mscope_params/cs_mm :  {2.7}
[CPU: 242.0 MB]      mscope_params/total_dose_e_per_A2 :  {53.3}
[CPU: 242.0 MB]             mscope_params/exp_group_id :  {1}
[CPU: 242.0 MB]              mscope_params/phase_plate :  {0}
[CPU: 242.0 MB]                mscope_params/neg_stain :  {0}
[CPU: 242.0 MB]                     movie_blob/psize_A :  {1.012}
[CPU: 242.0 MB]                       movie_blob/shape :  [  60 3710 3838]
[CPU: 242.0 MB]           movie_blob/is_gain_corrected :  {0}

I’m using 3.2.0 version of cryosparc & 1.3.1-Cuda102 version, nvcc 10.2.89. All jobs are run through the cryosparc interface.

Dear @marinegor,

Thank you for reporting. We have recently discovered this bug with the micrograph shapes, and a fix will be out in the next version.


1 Like