summaryrefslogtreecommitdiffstats
path: root/src/Python
diff options
context:
space:
mode:
authorEdoardo Pasca <edo.paskino@gmail.com>2017-11-01 16:32:08 +0000
committerEdoardo Pasca <edo.paskino@gmail.com>2018-01-19 14:26:06 +0000
commitabf93b6444b5dbba88de7489a56a6f30c7d687d8 (patch)
treeb15687e12ddd3414a6aab841330b706e9260a7b6 /src/Python
parentba360d14e7b501e2d9c4fd52f88ab604e126578f (diff)
downloadregularization-abf93b6444b5dbba88de7489a56a6f30c7d687d8.tar.gz
regularization-abf93b6444b5dbba88de7489a56a6f30c7d687d8.tar.bz2
regularization-abf93b6444b5dbba88de7489a56a6f30c7d687d8.tar.xz
regularization-abf93b6444b5dbba88de7489a56a6f30c7d687d8.zip
Initial demo as Demo_RealData3D_Parallel.m
Diffstat (limited to 'src/Python')
-rw-r--r--src/Python/demo/demo_dendrites.py138
1 files changed, 138 insertions, 0 deletions
diff --git a/src/Python/demo/demo_dendrites.py b/src/Python/demo/demo_dendrites.py
new file mode 100644
index 0000000..528702c
--- /dev/null
+++ b/src/Python/demo/demo_dendrites.py
@@ -0,0 +1,138 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Wed Aug 23 16:34:49 2017
+
+@author: ofn77899
+Based on DemoRD2.m
+"""
+
+import h5py
+import numpy
+
+from ccpi.reconstruction.FISTAReconstructor import FISTAReconstructor
+import astra
+import matplotlib.pyplot as plt
+from ccpi.imaging.Regularizer import Regularizer
+from ccpi.reconstruction.AstraDevice import AstraDevice
+from ccpi.reconstruction.DeviceModel import DeviceModel
+
+def RMSE(signal1, signal2):
+ '''RMSE Root Mean Squared Error'''
+ if numpy.shape(signal1) == numpy.shape(signal2):
+ err = (signal1 - signal2)
+ err = numpy.sum( err * err )/numpy.size(signal1); # MSE
+ err = sqrt(err); # RMSE
+ return err
+ else:
+ raise Exception('Input signals must have the same shape')
+
+filename = r'/home/ofn77899/Reconstruction/CCPi-FISTA_Reconstruction/demos/DendrData.h5'
+nx = h5py.File(filename, "r")
+#getEntry(nx, '/')
+# I have exported the entries as children of /
+entries = [entry for entry in nx['/'].keys()]
+print (entries)
+
+Sino3D = numpy.asarray(nx.get('/Sino3D'), dtype="float32")
+Weights3D = numpy.asarray(nx.get('/Weights3D'), dtype="float32")
+angSize = numpy.asarray(nx.get('/angSize'), dtype=int)[0]
+angles_rad = numpy.asarray(nx.get('/angles_rad'), dtype="float32")
+recon_size = numpy.asarray(nx.get('/recon_size'), dtype=int)[0]
+size_det = numpy.asarray(nx.get('/size_det'), dtype=int)[0]
+slices_tot = numpy.asarray(nx.get('/slices_tot'), dtype=int)[0]
+
+Z_slices = 20
+det_row_count = Z_slices
+# next definition is just for consistency of naming
+det_col_count = size_det
+
+detectorSpacingX = 1.0
+detectorSpacingY = detectorSpacingX
+
+
+proj_geom = astra.creators.create_proj_geom('parallel3d',
+ detectorSpacingX,
+ detectorSpacingY,
+ det_row_count,
+ det_col_count,
+ angles_rad)
+
+#vol_geom = astra_create_vol_geom(recon_size,recon_size,Z_slices);
+image_size_x = recon_size
+image_size_y = recon_size
+image_size_z = Z_slices
+vol_geom = astra.creators.create_vol_geom( image_size_x,
+ image_size_y,
+ image_size_z)
+
+
+## Create a Acquisition Device Model
+## Must specify some parameters of the acquisition:
+
+astradevice = AstraDevice(
+ DeviceModel.DeviceType.PARALLEL3D.value,
+ [det_row_count , det_col_count ,
+ detectorSpacingX, detectorSpacingY ,
+ angles_rad
+ ],
+ [ image_size_x, image_size_y, image_size_z ] )
+
+fistaRecon = FISTAReconstructor(proj_geom,
+ vol_geom,
+ Sino3D ,
+ weights=Weights3D,
+ device=astradevice,
+ Lipschitz_constant = 767893952.0,
+ subsets = 8)
+
+print("Reconstruction using FISTA-OS-PWLS without regularization...")
+fistaRecon.setParameter(number_of_iterations = 18)
+
+### adjust the regularization parameter
+##lc = fistaRecon.getParameter('Lipschitz_constant')
+##fistaRecon.getParameter('regularizer')\
+## .setParameter(regularization_parameter=5e6/lc)
+fistaRecon.use_device = True
+if False:
+ fistaRecon.prepareForIteration()
+ X = fistaRecon.iterate(numpy.load("../test/X.npy"))
+ numpy.save("FISTA-OS-PWLS.npy",X)
+
+## setup a regularizer algorithm
+regul = Regularizer(Regularizer.Algorithm.FGP_TV)
+regul.setParameter(regularization_parameter=5e6,
+ number_of_iterations=50,
+ tolerance_constant=1e-4,
+ TV_penalty=Regularizer.TotalVariationPenalty.isotropic)
+if False:
+ # adjust the regularization parameter
+ lc = fistaRecon.getParameter('Lipschitz_constant')
+ regul.setParameter(regularization_parameter=5e6/lc)
+ fistaRecon.setParameter(regularizer=regul)
+ fistaRecon.prepareForIteration()
+ X = fistaRecon.iterate(numpy.load("../test/X.npy"))
+ numpy.save("FISTA-OS-PWLS-TV.npy",X)
+
+if False:
+ # adjust the regularization parameter
+ lc = fistaRecon.getParameter('Lipschitz_constant')
+ regul.setParameter(regularization_parameter=5e6/lc)
+ fistaRecon.setParameter(regularizer=regul)
+ fistaRecon.setParameter(ring_lambda_R_L1=0.002, ring_alpha=21)
+ fistaRecon.prepareForIteration()
+ X = fistaRecon.iterate(numpy.load("../test/X.npy"))
+ numpy.save("FISTA-OS-GH-TV.npy",X)
+
+if True:
+ # adjust the regularization parameter
+ lc = fistaRecon.getParameter('Lipschitz_constant')
+ regul.setParameter(
+ algorithm=Regularizer.Algorithm.TGV_PD,
+ regularization_parameter=0.5/lc,
+ number_of_iterations=5)
+ fistaRecon.setParameter(regularizer=regul)
+ fistaRecon.setParameter(ring_lambda_R_L1=0.002, ring_alpha=21)
+ fistaRecon.prepareForIteration()
+ X = fistaRecon.iterate(numpy.load("../test/X.npy"))
+ numpy.save("FISTA-OS-GH-TGV.npy",X)
+