Create new task schema for iteratively sampling orientations from ODF in MTEX

Sam Engel already uses this method, but not as a Matflow task.

NB: Reported that this method converges rapidly on the origin model despite being more computationally heavy than simple random sampling, so is of a lot of interest.

Guy has a notebook which compares the two methods.

@Gbowker

@Gbowker would you be able to send me something to get me started with this? Thanks

Hi Gerard, Here is the script and input files I have which does this in MTEX:

texture sampling.zip

We just need to write the task_schema and workflow to accompany it:
Inputs:

.crc filepaths
corrections (one set of three angles for each .crc file)
Outputs:
combined_texture.png pole figures
sampled_texture.txt
sampled_texture.png pole figures

The script uses an older version of MTEX (5.3) but should work for a later version with small tweaks.
I needed to use MTEX 5.3 due to the requirements of the industrial partners.
Let me know if this is enough to get you started!

Thanks Guy. I've tried to run the script with MTEX 5.3 and MATLAB R2024b, but get an error:

47D.crc
    89    88   135

Error using loadEBSD_crc (line 30)
EBSD format 'CRC' does not match the data
 file: '47D.crc'

Error in EBSD.load (line 86)
  ebsd{k} = feval(['loadEBSD_',char(interface)],...
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in EBSD2model_varargs>loadmap_correct (line 129)
    ebsdmap = EBSD.load(filepath,CS,'interface','crc','convertSpatial2EulerReferenceFrame', 'Bunge');
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in EBSD2model_varargs (line 27)
    corrected_maps{m} = loadmap_correct(maps(m), CS, corrections(m,:));
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Would you mind checking the files in the zip archive? Thanks

Hi Gerard, Matlab just needs absolute paths rather than relative ones. So you need to include your pwd in the path for each .crc file.

Thanks Guy. I've tried that but I get the same error, only this time will the full rather than absolute path. It's complaining that the EBSD format doesn't match the data, rather than not being able to find the file.

I met with Guy and now have some new files which ran successfully.

Notes from our chat:

Don't want ebsd_maps.txt as input for the matflow task. Want list of (any number of) file paths, and corrections for each path (list of three correction values). Also need euler_degrees: true/false as input. mtex expects degrees.
Keep two versions of mtex in the matflow task: 5.3 and latest.
New files contain three datasets. Test all three run with MATLAB before moving on.
Get code working with both MTEX 5.3 and MTEX 6.0.0

For the MatFlow tasks, I think the existing MATLAB script does a few things which should be separated into distinct tasks:

Save ebsd orientations from combined maps
Plot pole figures for the ODF (task already exists in MatFlow)
Create ODF from orientations, then save texture sampled using MCMC
Plot pole figures for sampled texture (task already exists in MatFlow)

The implications for the above are that

the orientations needs to be saved after the first task as an HDF file as the required input format for the plot_pole_figures task
the ODF should be generated from orientations for the MCMC sampling

Implementing the first task in MatFlow has revealed #576.
I've been looking to reuse existing code (get_EBSD_orientations_from_CRC_file) to load the CRC files, but I'm unclear whether this will work for the purposes required here as we need to do three rotations. When I try I get this error about types:

34      corrected_maps{m} = get_EBSD_orientations_from_CRC_file(ebsd_map_path, ...
Error using MException
MESSAGE input must be a character vector or string scalar.

Error in interfaceError (line 28)
e = MException('mtex:wrongInterface',...
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in loadEBSD_crc (line 32)
  interfaceError(fname);
  ^^^^^^^^^^^^^^^^^^^^^
Error in Novelis_EBSD2model_varargs>get_EBSD_orientations_from_CRC_file (line 311)
    ebsd = loadEBSD_crc(CRCFilePath, referenceFrameTransformation{:});
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in Novelis_EBSD2model_varargs (line 34)
    corrected_maps{m} = get_EBSD_orientations_from_CRC_file(ebsd_map_path, ...

I'd confused some of the required input formats around cells, but the final problem is that the MTEX rotation function looks only to be able to do one rotation at a time, so we cannot pass in a vector of 3 euler angles and use the MatFlow function.
So use Guy's code instead, and Adam is looking into #576.

#576 has now been fixed.

Great, thankyou - how close are you to getting the MCMC task finished now?

Making steady progress. I could do with another chat next week if you're around on Monday?
~~It wasn't EBSD.orientations that we needed to export to HDF, rather orientations sampled from the ODF. So I think I've done that, but can't get the pole figure quite right.~~
Other tasks do that, but we need to save the actual orienations rather than a sample from an ODF.

I'm not sure whether I've captured all that is needed for the pole figures plot. Your script contains

CS = crystalSymmetry('m-3m', [4.04 4.04 4.04], 'mineral', 'Aluminum', 'color', [0 0.8 1]);

and I'm not sure whether the matflow input parameter crystal_symmetry: cubic is enough? The help/notation isn't super clear, but cubic looks to equate to the first term m-3m.

@Gbowker

Hi @Gbowker ,
I think this is ready for you to test now. If you pull this branch, there's one thing you'd need to change in the demo workflow, which is the path to the CRC files under

tasks:
- schema: combine_orientations_from_EBSD_maps
  inputs:
    CRC_file_paths:

I'll add some demo workflow files once you're happy with the workflow.

Fantastic Gerard will give that a go later today! :)
Thanks!

Great! I'm still looking at getting the pole figure plots the same as you made with MATLAB so watch this space 🤞

Debugging info for me for later:
The same MTEXprefs value is set in MATLAB and via MatFlow for pfAnnotations.
In MATLAB, this draws the XYZ labels, but in MatFlow it doesn't. Why?
In MATLAB, adding manual annotations draws on top the existing XYZ ones, which is messy.
In MatFlow, there are no existing annotations so it works out ok.

Maybe look at startup_mtex script and compare with Matflow matlab_env.

Navigation Menu

Repository navigation

Create new task schema for iteratively sampling orientations from ODF in MTEX #567

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Create new task schema for iteratively sampling orientations from ODF in MTEX #567

Description

Activity

dkfellows commented on Sep 1, 2026

gcapes commented on Sep 1, 2026

gcapes commented on Sep 7, 2026

Gbowker commented on Sep 7, 2026

gcapes commented on Sep 8, 2026

Gbowker commented on Sep 8, 2026

gcapes commented on Sep 8, 2026

gcapes commented on Sep 15, 2026

gcapes commented on Sep 17, 2026

gcapes commented on Sep 23, 2026

gcapes commented on Sep 24, 2026

gcapes commented on Sep 24, 2026

Gbowker commented on Sep 24, 2026

gcapes commented on Sep 24, 2026

gcapes commented on Sep 28, 2026

gcapes commented on Oct 1, 2026

Gbowker commented on Oct 1, 2026

gcapes commented on Oct 1, 2026

gcapes commented on Oct 1, 2026

gcapes commented on Oct 8, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions