MPSlib: Using masks

This example demonstrates how to use a spatial mask to restrict simulation to specific regions of the grid, and how to combine two simulations using different training images in different masked regions.

import mpslib as mps
import matplotlib.pyplot as plt
import numpy as np
import copy
from numpy import squeeze

Load training images

# TI1: Strebelle (original orientation)
TI1, TI_filename1 = mps.trainingimages.strebelle(di=1)
TI1 = np.swapaxes(TI1, 0, 1)
# TI2: Strebelle (rotated and coarsened)
TI2, TI_filename2 = mps.trainingimages.strebelle(di=2)
plt.figure(1)
plt.subplot(1, 2, 1)
plt.imshow(np.transpose(TI1[:, :, 0]))
plt.title('Training image 1')
plt.subplot(1, 2, 2)
plt.imshow(np.transpose(TI2[:, :, 0]))
plt.title('Training image 2')
plt.show()

Setup MPS object

grid_size = np.array([350, 200, 1])
O = mps.mpslib(method='mps_snesim_tree',
               n_real=1, verbose_level=-1)
#O = mps.mpslib(method='mps_genesim',
#               n_real=1, verbose_level=-1)
O.par['debug_level'] = -1
O.par['n_cond'] = 49
O.par['simulation_grid_size'] = grid_size

# Make sure no unwanted hard/soft data files are being used
O.delete_local_files()

Define masks

d_mask1 = np.zeros([grid_size[0], grid_size[1], grid_size[2]])
d_mask1[80:150, 100:180] = 1
d_mask1[0:40, 80:150] = 1
d_mask2 = 1 - d_mask1
mask_fnam1 = 'mask_01.dat'
mask_fnam2 = 'mask_02.dat'
mps.eas.write_mat(d_mask1, mask_fnam1)
mps.eas.write_mat(d_mask2, mask_fnam2)

{'dim': {'nx': 350, 'ny': 200, 'nz': 1},
 'n_cols': 1,
 'title': '350 200 1',
 'header': ['Header'],
 'D': array([1., 1., 1., ..., 1., 1., 1.], shape=(70000,))}

plt.figure(2)
plt.subplot(121)
plt.imshow(np.transpose(np.squeeze(d_mask1)))
plt.title('Mask 1')
plt.subplot(122)
plt.imshow(np.transpose(np.squeeze(d_mask2)))
plt.title('Mask 2')
plt.show()

Simulate in region 1 (mask 1)

O1 = copy.deepcopy(O)
O1.delete_hard_data()
O1.par['mask_fnam'] = mask_fnam1
O1.ti = TI1
O1.run()

True

plt.figure(3)
O1.plot_reals()

Extract hard data from simulation in region 1

d_hard = O1.hard_data_from_sim()

Simulate in region 2 (mask 2), conditioned on region 1

d_hard = O1.hard_data_from_sim()
O2 = copy.deepcopy(O)
O2.parameter_filename = 'mps_mask2.par'
O2.par['mask_fnam'] = mask_fnam2
O2.ti = TI2
O2.delete_hard_data()
O2.d_hard = d_hard
O2.run()

True

plt.figure(4)
O2.plot_reals()