2D Classification PCA failing

[ratolon]

I have the following set of data:

• 2.3M particles extracted with Xmipp
• 4A/px 74x74px
• 2D PCA "create_classes", 150 classes, max resolution 8.0 and 40k particles for training

I get the following error in step pcaTraining:

00001: Traceback (most recent call last):
00002: File "/home/miceta/software/scipion/software/em/xmipp/bin/xmipp_classify_pca_train", line 94, in
00003: freq_band = precalculateBands(nBand, dim, sampling, maxRes, minRes)
00004: File "/home/miceta/software/scipion/software/em/xmipp/bin/xmipp_classify_pca_train", line 39, in precalculateBands
00005: freq_band[y][x] = torch.floor(w*factor)
00006: IndexError: index 37 is out of bounds for dimension 0 with size 37
00007: Traceback (most recent call last):
00008: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/protocol/protocol.py", line 207, in run
00009: self._run()
00010: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/protocol/protocol.py", line 257, in _run
00011: resultFiles = self._runFunc()
00012: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/protocol/protocol.py", line 253, in _runFunc
00013: return self._func(*self._args)
00014: File "/home/miceta/software/xmipp/src/scipion-em-xmipp/xmipp3/protocols/protocol_classify_pca.py", line 236, in pcaTraining
00015: self.runJob("xmipp_classify_pca_train", args, numberOfMpi=1, env=env)
00016: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/protocol/protocol.py", line 1545, in runJob
00017: self._stepsExecutor.runJob(self._log, program, arguments, **kwargs)
00018: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/protocol/executor.py", line 68, in runJob
00019: process.runJob(log, programName, params,
00020: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/utils/process.py", line 56, in runJob
00021: return runCommand(command, env=env, cwd=cwd, executable=executable)
00022: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/site-packages/pyworkflow/utils/process.py", line 71, in runCommand
00023: check_call(command, shell=True, stdout=sys.stdout, stderr=sys.stderr,
00024: File "/home/miceta/software/miniforge3/envs/scipion3/lib/python3.8/subprocess.py", line 364, in check_call
00025: raise CalledProcessError(retcode, cmd)
00026: subprocess.CalledProcessError: Command ' xmipp_classify_pca_train -i Runs/006886_XmippProtClassifyPca/tmp/images.mrc -s 4.0 -hr 8.0 -lr 530 -p 0.75 -t 40000 -o Runs/006886_XmippProtClassifyPca/extra/train_pca --batchPCA' returned non-zero exit status 1.
00027: Protocol failed: Command ' xmipp_classify_pca_train -i Runs/006886_XmippProtClassifyPca/tmp/images.mrc -s 4.0 -hr 8.0 -lr 530 -p 0.75 -t 40000 -o Runs/006886_XmippProtClassifyPca/extra/train_pca --batchPCA' returned non-zero exit status 1.

[ratolon]

The problem is related to the sampling rate and the maximum resolution in the protocol form. For instance, 4A/px makes it fail if max resolution is equal or less than Nyquist.

Lo dejamos para el hunting o para ponerlo antes de la release. Thank you very much Erney!!