-
Notifications
You must be signed in to change notification settings - Fork 8
/
local_analysis_vis.py
250 lines (206 loc) · 9.66 KB
/
local_analysis_vis.py
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
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from mpl_toolkits.axes_grid1 import ImageGrid
import matplotlib.gridspec as gridspec
from PIL import Image
import numpy as np
import os
import argparse
import re
import shutil
classname_dict = dict()
classname_dict[0] = "circumscribed"
classname_dict[1] = "indistinct"
classname_dict[2] = "spiculated"
def main():
# get dir
parser = argparse.ArgumentParser()
parser.add_argument('-local_analysis_directory', nargs=1, type=str, default='0')
args = parser.parse_args()
source_dir = args.local_analysis_directory[0]
os.makedirs(os.path.join(source_dir, 'visualizations_of_expl'), exist_ok=True)
pred, truth = read_local_analysis_log(os.path.join(source_dir + 'local_analysis.log'))
anno_opts_cen = dict(xy=(0.4, 0.5), xycoords='axes fraction',
va='center', ha='center')
anno_opts_symb = dict(xy=(1, 0.5), xycoords='axes fraction',
va='center', ha='center')
anno_opts_sum = dict(xy=(0, -0.1), xycoords='axes fraction',
va='center', ha='left')
###### all classes, one expl
fig = plt.figure(constrained_layout=False)
fig.set_size_inches(28, 12)
ncols, nrows = 7, 3
spec = gridspec.GridSpec(ncols=ncols, nrows=nrows, figure=fig)
f_axes = []
for row in range(nrows):
f_axes.append([])
for col in range(ncols):
f_axes[-1].append(fig.add_subplot(spec[row, col]))
plt.rcParams.update({'font.size': 14})
for ax_num, ax in enumerate(f_axes[0]):
if ax_num == 0:
ax.set_title("Test image", fontdict=None, loc='left', color = "k")
elif ax_num == 1:
ax.set_title("Test image activation\nby prototype", fontdict=None, loc='left', color = "k")
elif ax_num == 2:
ax.set_title("Prototype", fontdict=None, loc='left', color = "k")
elif ax_num == 3:
ax.set_title("Self-activation of\nprototype", fontdict=None, loc='left', color = "k")
elif ax_num == 4:
ax.set_title("Similarity score", fontdict=None, loc='left', color = "k")
elif ax_num == 5:
ax.set_title("Class connection", fontdict=None, loc='left', color = "k")
elif ax_num == 6:
ax.set_title("Contribution", fontdict=None, loc='left', color = "k")
else:
pass
plt.rcParams.update({'font.size': 22})
for ax in [f_axes[r][4] for r in range(nrows)]:
ax.annotate('x', **anno_opts_symb)
for ax in [f_axes[r][5] for r in range(nrows)]:
ax.annotate('=', **anno_opts_symb)
# get and plot data from source directory
orig_img = Image.open(os.path.join(source_dir + 'original_img.png'))
for ax in [f_axes[r][0] for r in range(nrows)]:
ax.imshow(orig_img)
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
top_p_dir = os.path.join(source_dir + 'most_activated_prototypes')
for top_p in range(3):
# put info in place
p_info_file = open(os.path.join(top_p_dir, f'top-{top_p+1}_activated_prototype.txt'), 'r')
sim_score, cc_dict, class_str, top_cc_str = read_info(p_info_file)
p_info_file.close()
for ax in [f_axes[top_p][4]]:
ax.annotate(sim_score, **anno_opts_cen)
ax.set_axis_off()
for ax in [f_axes[top_p][5]]:
ax.annotate(top_cc_str + "\n" + class_str, **anno_opts_cen)
ax.set_axis_off()
for ax in [f_axes[top_p][6]]:
tc = float(top_cc_str) * float(sim_score)
ax.annotate('{0:.3f}'.format(tc) + "\n" + class_str, **anno_opts_cen)
ax.set_axis_off()
# put images in place
p_img = Image.open(os.path.join(top_p_dir, f'top-{top_p+1}_activated_prototype_full_size.png'))
for ax in [f_axes[top_p][2]]:
ax.imshow(p_img)
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
p_act_img = Image.open(os.path.join(top_p_dir, f'top-{top_p+1}_activated_prototype_self_act.png'))
for ax in [f_axes[top_p][3]]:
ax.imshow(p_act_img)
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
act_img = Image.open(os.path.join(top_p_dir, f'prototype_activation_map_by_top-{top_p+1}_prototype_normed.png'))
for ax in [f_axes[top_p][1]]:
ax.imshow(act_img)
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
#summary
f_axes[2][4].annotate(f"This {classname_dict[int(truth)]} lesion is classified as {classname_dict[int(pred)]}.", **anno_opts_sum)
save_loc1 = os.path.join(source_dir, 'visualizations_of_expl') + f'/all_class.png'
plt.savefig(save_loc1, bbox_inches='tight', pad_inches=0)
os.makedirs('./visualizations_of_expl/', exist_ok=True)
save_loc2 = './visualizations_of_expl/' + str(source_dir.replace('/', '__'))[len('__usr__xtmp__IAIABL__saved_models__0129_pushonall_topkk=9_fa=0.001_random=4__pruned_prototypes_epoch50_k6_pt3__'):] + f'all_class.png'
shutil.copy2(save_loc1, save_loc2)
print(f"Saved in {save_loc2}")
return
def read_local_analysis_log(file_loc):
log_file = open(file_loc, 'r')
for _ in range(30):
line = log_file.readline()
if line[0:len("Predicted: ")] == "Predicted: ":
pred = line[len("Predicted: "):]
elif line[0:len("Actual: ")] == "Actual: ":
actual = line[len("Actual: "):]
# pred = log_file.readline()[len("Predicted: "):]
# actual = log_file.readline()[len("Actual: "):]
log_file.close()
return pred, actual
def read_info(info_file, per_class=False):
sim_score_line = info_file.readline()
connection_line = info_file.readline()
proto_index_line = info_file.readline()
cc_0_line = info_file.readline()
cc_1_line = info_file.readline()
cc_2_line = info_file.readline()
sim_score = sim_score_line[len("similarity: "):-1]
if per_class:
cc = connection_line[len('last layer connection: '):-1]
else:
cc = connection_line[len('last layer connection with predicted class: '):-1]
circ_cc_str = cc_0_line[len('proto connection to class 0:tensor('):-(len(", device='cuda:0', grad_fn=<SelectBackward>)")+1)]
circ_cc = float(circ_cc_str)
indst_cc_str = cc_1_line[len('proto connection to class 1:tensor('):-(len(", device='cuda:0', grad_fn=<SelectBackward>)")+1)]
indst_cc = float(indst_cc_str)
spic_cc_str = cc_2_line[len('proto connection to class 2:tensor('):-(len(", device='cuda:0', grad_fn=<SelectBackward>)")+1)]
spic_cc = float(spic_cc_str)
cc_dict = dict()
cc_dict[0] = circ_cc
cc_dict[1] = indst_cc
cc_dict[2] = spic_cc
class_of_p = max(cc_dict, key=lambda k: cc_dict[k])
top_cc = cc_dict[class_of_p]
class_str = classname_dict[class_of_p]
if class_of_p == 0:
top_cc_str = circ_cc_str
elif class_of_p == 1:
top_cc_str = indst_cc_str
elif class_of_p == 2:
top_cc_str = spic_cc_str
else:
print("Error. The maximum value class is not found.")
return sim_score, cc_dict, class_str, top_cc_str
def test():
im = Image.open('./visualizations_of_expl/' + 'original_img.png')
fig = plt.figure(constrained_layout=False)
fig.set_size_inches(28, 12)
ncols, nrows = 7, 3
spec = gridspec.GridSpec(ncols=ncols, nrows=nrows, figure=fig)
f_axes = []
for row in range(nrows):
f_axes.append([])
for col in range(ncols):
f_axes[-1].append(fig.add_subplot(spec[row, col]))
plt.rcParams.update({'font.size': 15})
for ax_num, ax in enumerate(f_axes[0]):
if ax_num == 0:
ax.set_title("Test image", fontdict=None, loc='left', color = "k")
elif ax_num == 1:
ax.set_title("Test image activation by prototype", fontdict=None, loc='left', color = "k")
elif ax_num == 2:
ax.set_title("Prototype", fontdict=None, loc='left', color = "k")
elif ax_num == 3:
ax.set_title("Self-activation of prototype", fontdict=None, loc='left', color = "k")
elif ax_num == 4:
ax.set_title("Similarity score", fontdict=None, loc='left', color = "k")
elif ax_num == 5:
ax.set_title("Class connection", fontdict=None, loc='left', color = "k")
elif ax_num == 6:
ax.set_title("Contribution", fontdict=None, loc='left', color = "k")
else:
pass
plt.rcParams.update({'font.size': 22})
for ax in [f_axes[r][0] for r in range(nrows)]:
ax.imshow(im)
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
anno_opts = dict(xy=(0.4, 0.5), xycoords='axes fraction',
va='center', ha='center')
anno_opts_symb = dict(xy=(1, 0.5), xycoords='axes fraction',
va='center', ha='center')
for ax in [f_axes[r][s] for r in range(nrows) for s in range(4,7)]:
ax.annotate('Number', **anno_opts)
ax.set_axis_off()
for ax in [f_axes[r][4] for r in range(nrows)]:
ax.annotate('x', **anno_opts_symb)
for ax in [f_axes[r][5] for r in range(nrows)]:
ax.annotate('=', **anno_opts_symb)
os.makedirs('./visualizations_of_expl/', exist_ok=True)
plt.savefig('./visualizations_of_expl/' + 'test.png')
# Refs: https://stackoverflow.com/questions/40846492/how-to-add-text-to-each-image-using-imagegrid
# https://stackoverflow.com/questions/41793931/plotting-images-side-by-side-using-matplotlib
if __name__ == "__main__":
main()