Config file generation
Example config file below:
### config file for LaueNeuralNetwork
## comments
[CPU]
n_cpu = 16
[GLOBAL_DIRECTORY]
prefix = _final
## directory where all training related data and results will be saved
main_directory = C:UserspurushotDesktoppattern_matchingexperimentalGUIv0latest_version
[MATERIAL]
## same material key as lauetools (see Lauetools.dictlauetools.py for complete key)
## as of now symmetry can be cubic, hexagonal, orthorhombic, tetragonal, trigonal, monoclinic, triclinic
material = In2Bi
symmetry = hexagonal
space_group = 194
#general_diffraction_rules will filter out the forbidden spots
#(slow, keep it false, as they will be eliminated at a later stage;
#useful if system has low RAM to eliminate some possibilities before)
general_diffraction_rules = false
## if second phase is present, else none
material1 = In_epsilon
symmetry1 = tetragonal
space_group1 = 139
general_diffraction_rules1 = false
[DETECTOR]
## path to detector calibration file (.det)
## Max and Min energy to be used for generating training dataset, as well as for calcualting matching rate
detectorfile = C:UserspurushotDesktopIn_JSMcalib.det
# or use as below if you have values
#detectorfile = user_input
#five detector parameters+pixelsize+image_dimensions+ccd_label
#params = 79.51900,1951.6300,1858.1500,0.3480000,0.4560000,0.03670000,4036,4032,sCMOS_16M
emax = 22
emin = 5
[TRAINING]
## classes_with_frequency_to_remove: HKL class with less appearance than specified will be ignored in output
## desired_classes_output : can be all or an integer: to limit the number of output classes
## max_HKL_index : can be auto or integer: Maximum index of HKL to build output classes
## max_nb_grains : Maximum number of grains to simulate per lauepattern
####### Material 0
classes_with_frequency_to_remove = 500
desired_classes_output = all
max_HKL_index = 5
max_nb_grains = 1
####### Material 1
## HKL class with less appearance than specified will be ignored in output
classes_with_frequency_to_remove1 = 500
desired_classes_output1 = all
max_HKL_index1 = 5
max_nb_grains1 = 1
## Max number of simulations per number of grains
## Include single crystal misorientation (1 deg) data in training
## Maximum angular distance to probe (in deg)
## step size in angular distribution to discretize (in deg)
## batch size and epochs for training
max_simulations = 500
include_small_misorientation = false
misorientation_angle = 1
angular_distance = 120
step_size = 0.1
batch_size = 50
epochs = 5
[PREDICTION]
# model_weight_file: if none, it will select by default the latest H5 weight file, else provide a specific model
# softmax_threshold_global: thresholding to limit the predicted spots search zone
# mr_threshold_global: thresholding to ignore all matricies less than the MR threshold
# cap_matchrate: any UB matrix providing MR less than this will be ignored (between 0-1)
# coeff: should be same as cap_matchrate but this is for try previous UB matrix
# coeff_overlap: coefficient to limit the overlapping between spots; if more than this, new solution will be computed
# mode_spotCycle: How to cycle through predicted spots (slow or graphmode) ##slow is more thorough but slow as the name suggests
##use_previous true for few crystal and prefered texture case, otherwise time consuming; advised for single phase alone
UB_matrix_to_detect = 2
matrix_tolerance = 0.6
matrix_tolerance1 = 0.6
cap_matchrate = 0.10
material0_limit = 1
material1_limit = 1
### no need to change the settings below unless the neural network does not give satisfactory results
mode_spotCycle = graphmode
model_weight_file = none
softmax_threshold_global = 0.85
mr_threshold_global = 0.80
coeff = 0.3
coeff_overlap = 0.3
#true for few crystal and prefered texture case, otherwise time consuming; advised for single phase alone
#true also if using single CPU mode to make indexation go faster
use_previous = false
[EXPERIMENT]
experiment_directory = user_path_to_experimental_folder
experiment_file_prefix = file_prefix_without_filenumbers
image_grid_x = 51
image_grid_y = 51
[PEAKSEARCH]
# mode: either LaueTools or skimage
# Two modes of peaksearch is possible
# For skimage, keep intenisty_threshold somewhere between 2 - 10
intensity_threshold = 100
boxsize = 15
fit_peaks_gaussian = 1
FitPixelDev = 15
mode = LaueTools
[STRAINCALCULATION]
# strain computation with multi step refinement for material0 and material1
strain_compute = true
tolerance_strain_refinement = 0.6,0.5,0.4,0.3,0.2
tolerance_strain_refinement1 = 0.6,0.5,0.4,0.3,0.2
[CALLER]
# some additional settings to change hardcoded variables in the code
residues_threshold=0.25
nb_spots_global_threshold=8
option_global = v2
#use_om_user provides the possibility to provide your own path with text file contatining orientations,
#if true, no prediction will be done with neural network and we just index with user defined orientations
use_om_user = false
#first n number of intense spots to predict the hkl for, for many UB detection keep a high number like 1000
nb_spots_consider = 200
[DEVELOPMENT]
# could be 1 or 2 or none in case of single phase (for pretty plots) or in case of substrate present
material_phase_always_present = none
#writes MTEX input file to be treated with MTEX for advance orientation analysis
write_MTEX_file = true