WignerAI/testprepim.py at master · samusake/WignerAI · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Wed May 22 15:25:37 2019

@author: herbert
"""
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
import skimage as sk
import json

import time

import tensorflow as tf
from tensorflow.keras import layers
from tensorflow import keras
from keras.utils import to_categorical
from keras.models import model_from_json
from tensorflow.keras.callbacks import ModelCheckpoint

from genSample import *
from model import *

from skimage.transform import iradon, iradon_sart
from skimage.transform import radon, rescale
from scipy.special import erf
#%%
stest=3060 #number of samples
nphi=42#45 #number of angleSteps

nxs=42
xmax=5
lxs=np.linspace(-xmax, xmax, nxs)
[xs, ys]=np.meshgrid(lxs,lxs);

phispace=np.linspace(0,180,nphi, endpoint=False)
[px, py]=np.meshgrid(lxs,phispace)
#%%
#stest=10

P=np.load('data/image_learn/P30608_images.npy')[(30608-3060):]
W=np.load('data/image_learn/W30608_images.npy')[(30608-3060):]

#%%
P_radon=np.zeros((stest, 60,nphi))
for i in range(0,stest):
    #P_radon[i]=generateP_radonofw(W[i],lxs,phispace)
    P_radon[i]=radon(W[i],theta=phispace, circle=False)
#%%
json_file = open('models/images42/ai_model.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
ai = keras.models.model_from_json(loaded_model_json)
# load weights into new model
ai.load_weights("models/images42/ai_checkpoint.h5")
print("Loaded model from disk")
#%%
with open('models/images42/ai_history.json') as json_file:
    history=json.load(json_file)

plt.semilogy(history['loss'][3:499])
plt.semilogy(history['val_loss'][3:499])

plt.title('Model loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['Train', 'Test'], loc='upper left')
plt.show()
#%%
reconstruction_fbp=np.zeros((stest, nxs,nphi))
start=time.time()
for i in range(0,stest):
    reconstruction_fbp[i] = iradon(P_radon[i], theta=phispace)
end=time.time()
error = reconstruction_fbp - W
print('FBP reconstruction error (3060 samples): %.3g' % np.sqrt(np.mean(error**2)))
print('FBP calculation duration (3060 samples): %.3g' % (end-start))
print('FBP mean calculation duration: %.3g' % ((end-start)/stest))
#sinogram=radon(W[i], theta=phispace, circle=True)

inputV=np.zeros((stest,60*nphi))
for i in range(0, len(P)):
    inputV[i]=P[i].flatten()


#predict for all
start=time.time()
wai_orig=ai.predict(inputV)
end=time.time()
wai=np.concatenate(wai_orig)
wai=np.reshape(wai, (stest,nxs,nxs))

error = wai - W


print('AI reconstruction error (3060 samples): %.3g' % np.sqrt(np.mean(error**2)))
print('AI prediction duration (3060 samples): %.3g' % (end-start))

start=time.time()
for i in range(0,stest):
    wai_orig=ai.predict(np.array([inputV[i]]))
end=time.time()
print('AI mean prediction duration: %.3g' % ((end-start)/stest))


#%%
#contour1=erf(np.linspace(-0.7,0.7,100)/np.sqrt(2))
#contour2=erf(np.linspace(-0.5,1,100)/np.sqrt(2*0.4))
k=55
fig, axs = plt.subplots(4,4, sharex=True)
for i in range(0,4):
    axs[0,i].contourf(P[i+k]/np.mean(P[i+k]),levels=15,extend='both')
    axs[1,i].imshow(W[i+k], cmap='gray')
    #axs[1,i].axis('equal')
    axs[2,i].imshow(wai[i+k], cmap='gray')
    #axs[2,i].axis('equal')
    axs[3,i].imshow(reconstruction_fbp[i+k], cmap='gray')
    #axs[3,i].axis('equal')


axs[0,0].set_ylabel('phi')
axs[1,0].set_ylabel('Y')
axs[2,0].set_ylabel('Y')
axs[3,0].set_ylabel('Y')

axs[3,0].set_xlabel('X')
axs[3,1].set_xlabel('X')
axs[3,2].set_xlabel('X')
axs[3,3].set_xlabel('X')

axs[0,1].title.set_text('Sinogram')
axs[1,1].title.set_text('Ground truth')
axs[2,1].title.set_text('AI prediction')
axs[3,1].title.set_text('FBP (skimage implementation)')

plt.show()