-
Notifications
You must be signed in to change notification settings - Fork 0
/
graphing.py
288 lines (245 loc) · 10.7 KB
/
graphing.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
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
import os
import sys
import cv2
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import argparse
from scipy.special.orthogonal import legendre
from preprocessing import readNoiseText
import numpy as np
import copy
def parseArgs():
parser = argparse.ArgumentParser()
parser.add_argument('--noise_file', type=str, default='noise.txt')
parser.add_argument('--correction_file', type=str, default='corrections.txt')
parser.add_argument('--result_dir', type=str, default='./results')
return parser.parse_args()
def graph_2(noise_dict, result_root):
col = plt.cm.cool([0.25,0.75,1])
for item in noise_dict.items():
mAP = {}
plt.figure()
plt.title(item[0])
x = item[1]
for r in item[1]:
file_path = os.path.join(result_root, 'retinaface', '%s_%0.6f_map.txt'%(item[0], r))
with open(file_path, 'r') as f:
contents = list(map(str.strip, f.readlines()))
for con in contents:
cat, val = con.split(',')
val = float(val)
if not cat in mAP:
mAP[cat] = [val]
else:
mAP[cat].append(val)
legend = list(mAP.keys())
for i, map_item in enumerate(mAP.items()):
y = map_item[1]
plt.plot(x,y, color=col[i])
if item[0] == 'speckle':
plt.gca().invert_xaxis()
plt.legend(legend)
plt.show()
def graph_1(noise_dict, result_root, models):
def sort_first(x):
return x[0][0]
eval_modes = ['easy_val', 'medium_val', 'hard_val']
for item in noise_dict.items():
avg = []
colors = cm.rainbow(np.linspace(0,1,len(models)*len(eval_modes)))
fig = plt.figure(figsize=(10,5.5))
x = item[1]
model_container = {i:[] for i in models}
for mod in models:
emd = {i:[] for i in eval_modes}
for r in item[1]:
file_path = os.path.join(result_root, mod, '%s_%0.6f_map.txt'%(item[0], r))
with open(file_path, 'r') as f:
contents = list(map(str.strip, f.readlines()))
for con in contents:
cat, val = con.split(',')
val = float(val)
emd[cat].append(val)
model_container[mod].append(emd)
plt.title(determineProperName(item[0]))
legend = []
plot_y = []
for m in model_container.items():
for _emd in m[1]:
for k,n in enumerate(_emd.items()):
plot_y.append((n[1], '%s %s'%(m[0], n[0])))
plot_y.sort(key=sort_first)
if len(avg) == 0:
avg = copy.copy(n[1])
else:
for j,nval in enumerate(n[1]):
avg[j] = avg[j] + nval
#plt.plot(x,n[1], color=colors[i])
for _n,_c in zip(plot_y, colors):
print(_n)
plt.plot(x, _n[0], color=_c)
legend.append(_n[1])
avg = [_avg/len(legend) for _avg in avg]
plt.plot(x,avg, color='black', linewidth=3)
legend.append('average')
# flip x-axis for speckle:
if item[0] == 'speckle':
plt.gca().invert_xaxis()
plt.ylabel('mAP (%)')
plt.xlabel(determineXLabel(item[0]))
plt.ylim([0,1])
plt.legend(legend, bbox_to_anchor=(1.05, 1), loc='upper left')
plt.tight_layout()
plt.savefig('./results/graphs/overview_%s_graph.png'%item[0])
def graph_3(noise_dict, correct_dict, result_root, models):
def sort_inner(x):
return x[0]
val_weights = [0.33,0.33,0.33]
# assemble all maps from result_root:
for correct in correct_dict.items():
for noise in noise_dict.items():
for mod in models:
fig = plt.figure(figsize=(10,5.5))
x = copy.copy(correct[1])
legend = []
colors = cm.rainbow(np.linspace(0,1,len(noise[1])+1))
noise_map = []
for noi in noise[1]:
legend.append('%s @ %f'%(determineProperName(noise[0]), noi))
mAP = []
# grab noise-only value to represent the baseline
default_path = os.path.join(result_root, mod, "%s_%0.6f_map.txt"%(noise[0], noi))
avg = 0
with open(default_path, 'r') as f:
contents = list(map(str.strip, f.readlines()))
for i, con in enumerate(contents):
cat, val = con.split(',')
val = float(val)
avg += val*val_weights[i]
mAP.append(avg)
for cor in correct[1]:
#print(correct)
_path = os.path.join(result_root, mod, "%s_%s_%0.6f_%0.6f_map.txt"%(noise[0], correct[0], noi, cor) )
print(_path)
with open(_path, 'r') as f:
contents = list(map(str.strip, f.readlines()))
avg = 0
for i, con in enumerate(contents):
cat, val = con.split(',')
val = float(val)
avg += val*val_weights[i]
mAP.append(avg)
noise_map.append(mAP)
# add the default on the y:
x.insert(0, 1)
# plot out:
for c, _noise in enumerate(noise_map):
print(x, _noise)
plt.plot(x, _noise, color=colors[c])
if correct[0] == 'median':
plt.xlabel('Kernel Size')
elif correct[0] == 'nplf':
plt.xlabel('Non-linear Mean Filter')
plt.ylabel('mAP (%)')
plt.title('Correcting %s with %s on %s'%(determineProperName(noise[0]), determineProperName(correct[0]), determineProperName(mod)))
plt.xticks(x)
plt.ylim([0,1])
plt.legend(legend, bbox_to_anchor=(1.05, 1), loc='upper left')
plt.tight_layout()
plt.savefig('./results/graphs/overview_%s_%s_%s.png'%(correct[0], noise[0], mod))
# for histogram equiliazation only:
def graph_4(noise_dict, result_root, models):
val_weights = [0.33,0.33,0.33]
colors = ['red', 'orange']
legend = ['Baseline', 'Histogram Equalization']
for i, noise in enumerate(noise_dict.items()):
for mod in models:
mAP = []
fig = plt.figure(figsize=(10,5.5))
x = noise[1]
baseline = {'easy_val':[], 'medium_val':[], 'hard_val':[], 'avg':[]}
for noi in noise[1]:
# grab noise-only value to represent the baseline
default_path = os.path.join(result_root, mod, "%s_%0.6f_map.txt"%(noise[0], noi))
print(default_path)
avg = 0
with open(default_path, 'r') as f:
contents = list(map(str.strip, f.readlines()))
for i, con in enumerate(contents):
cat, val = con.split(',')
val = float(val)
avg += val*val_weights[i]
#baseline[cat].append(val)
baseline['avg'].append(avg)
mAP.append(baseline)
line = {'easy_val':[], 'medium_val':[], 'hard_val':[], 'avg':[]}
for noi in noise[1]:
avg = 0
_path = os.path.join(result_root, mod, "%s_%s_%0.6f_%0.6f_map.txt"%(noise[0], 'he', noi, 0) )
with open(_path, 'r') as f:
contents = list(map(str.strip,f.readlines()))
for i, con in enumerate(contents):
cat, val = con.split(',')
val = float(val)
avg += val*val_weights[i]
line[cat].append(val)
line['avg'].append(avg)
mAP.append(line)
for j, m in enumerate(mAP):
for val in m.items():
if val[1] != 'avg':
pass
#plt.plot(val[1], color='cornflowerblue')
plt.plot(x, m['avg'], color=colors[j])
plt.legend(legend, bbox_to_anchor=(1.05, 1), loc='upper left')
plt.ylabel('mAP (%)')
plt.title('Correcting %s with %s on %s'%(determineProperName(noise[0]), 'Histogram Equalization', determineProperName(mod)))
plt.xticks(x)
plt.ylim([0,1])
plt.tight_layout()
#plt.show()
plt.savefig('./results/graphs/overview_%s_%s_%s.png'%('he', noise[0], mod))
def determineXLabel(input_str):
if input_str == 'gaussian_noise':
return 'Standard Deviation'
elif input_str == 'salt_pepper':
return 'Percentage of Image'
elif input_str == 'poisson':
return 'Lambda'
elif input_str == 'speckle':
return 'Uniform Low Value'
elif input_str == 'gamma':
return 'Gamma Shape'
def determineProperName(input_str):
if input_str == 'gaussian_noise':
return 'Gaussian Noise'
elif input_str == 'salt_pepper':
return 'Salt and Pepper'
elif input_str == 'poisson':
return 'Poisson'
elif input_str == 'speckle':
return 'Speckle'
elif input_str == 'median':
return 'Median'
elif input_str == 'gamma':
return 'Gamma'
elif input_str == 'he':
return 'Histogram Equalization'
elif input_str == 'dsfd':
return 'DSFD'
elif input_str == 'retinaface':
return 'RetinaFace'
else:
raise Exception('Invalid input_str specified!')
if __name__ == '__main__':
args = parseArgs()
with open(args.noise_file, 'r') as f:
contents = list(map(str.strip, f.readlines()))
noise_dict = readNoiseText(contents)
with open(args.correction_file, 'r') as f:
contents = list(map(str.strip, f.readlines()))
correction_list = readNoiseText(contents)
graph_2(noise_dict, args.result_dir)
graph_1(noise_dict, args.result_dir, ['retinaface', 'tinaface', 'dsfd'])
graph_3(noise_dict, correction_list, args.result_dir, ['retinaface', 'dsfd'])
graph_4(noise_dict, args.result_dir, ['retinaface', 'dsfd'])