/
auto_labeler.py
317 lines (235 loc) · 7.64 KB
/
auto_labeler.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
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
import os
import csv
import argparse
import PIL.Image
import coremltools
import random
import math
import time
parser = argparse.ArgumentParser(description='Use a folder of ML model classifiers to label a local unlabeled data set')
parser.add_argument('-m', '--modeldir', type=str, help='folder containing core ml models to use as labelers. Each image to label will be run through each model', default='./Models/Classifiers/Cleaned/', required=False)
parser.add_argument('-i', '--imagedir', type=str, help="folder containing unlabeled images to be labeled", default="./images", required=True)
parser.add_argument('-o', '--output', type=str, help="destination for labeled file containing multi labels", default="./labels", required=False)
parser.add_argument('-t', '--type', type=str, help="csv or html?", default="csv", required=False)
parser.add_argument('-pre', '--prefix', type=str, help="image url prefix, useful for adding a cloud storage provider URL for example", default="", required=False)
parser.add_argument('-l', '--limit', type=int, help="limit the number of images we label - useful for testing", default="1000000000000", required=False)
parser.add_argument('-r', '--random', type=bool, help="limit the number of images we label - useful for testing", default=False, required=False)
parser.add_argument('-p', '--probabilities', type=bool, help="report probabilities rather than predicted class label (html only)", default=False, required=False)
args = parser.parse_args()
start = time.time()
# load our models into our models array
dir_path = os.getcwd()
models_path = os.path.normpath( os.path.join(dir_path, args.modeldir) )
print('Loading Models from: ' + models_path)
models = []
modelfiles = os.listdir(models_path)
modelfiles.sort()
for filename in modelfiles:
if filename.endswith('.mlmodel'):
model_path = (os.path.join(models_path, filename))
if model_path:
model = coremltools.models.MLModel(model_path)
if model:
print('Loaded model ' + filename)
models.append(model)
else:
print('Unable to load model at ' + model_path)
else:
continue
end = time.time()
modeltime = end - start
print("")
print("Loading models took " + str(modeltime) + " seconds")
print("")
Height = 224 # use the correct input image height
Width = 224 # use the correct input image width
def load_image(path, resize_to=None):
try:
img = PIL.Image.open(path)
#verify apparently breaks the image!?
# img.verify()
except Exception:
print('Unable to load image' + path)
return None
if resize_to is not None:
try:
img = img.resize(resize_to, PIL.Image.ANTIALIAS)
except Exception:
print('Unable to resize image' + path)
return None
# ensure we pass our image as RGB - some images might be single channel or RGBA
if img.mode != 'RGB':
try:
img = img.convert(mode='RGB')
except Exception:
print('Unable to convert image to RGB' + path)
return None
return img
def html_header():
html_header = """
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
"http://www.w3.org/TR/html4/loose.dtd">
<html>
<head>
<title>Synopsis Data Set Auto Labeler Output</title>
<style>
body {
font-family: Arial, Sans-serif;
font-size: 10pt;
}
.masonry-layout {
column-count: 3;
column-gap: 0;
}
.masonry-layout__panel {
break-inside: avoid;
margin:5px;
padding: 5px;
}
.masonry-layout__panel-content {
padding: 10px;
border-radius: 10px;
border: solid 1px gray;
background-color: #ddd;
}
@media screen and (min-width: 600px) {
.masonry-layout {
column-count: 2;
}
}
@media screen and (min-width: 800px) {
.masonry-layout {
column-count: 3;
}
}
@media screen and (min-width: 1000px) {
.masonry-layout {
column-count: 4;
}
}
@media screen and (min-width: 1200px) {
.masonry-layout {
column-count: 5;
}
}
#bar {
height: 100%;
background-color: green;
}
#label {
}
</style
</head>
<body>
<div class="masonry-layout">
"""
return html_header
def html_entry_label(filepath, labels):
html_entry = """
<div class="masonry-layout__panel">
<div class="masonry-layout__panel-content" align="center">
<a href="file://{}"><img src="file://{}" width="100%" target="_blank"/></a><br/> {}
</div>
</div>
"""
# item 0 is the image name
del labels[0]
filepath = os.path.normpath( os.path.join(dir_path, filepath) )
return html_entry.format(filepath, filepath, '<br />'.join(labels) )
def html_entry_scores_table(items):
html = '<div style="max-height:300px; overflow-y:scroll">'
html += '<table style="width:100%;">'
for i in range(0, len(items)):
label = items[i][0]
score = items[i][1]
percentString = str( math.ceil( score * 10000.0) / 100.0 ) + '%'
html += '<tr>'
html += '<td>'
html += '<div id="bar" style="height:17px; width:' + percentString + '">'+ label + '</div>'
html += '</td>'
html += '<td>' + percentString + '</td>'
html += '</tr>'
html += '</table>'
html += '</div>'
return html
def html_entry_scores(filepath, scores):
html_entry = """
<div class="masonry-layout__panel">
<div class="masonry-layout__panel-content" align="center">
<a href="file://{}"><img src="file://{}" width="100%" target="_blank"/></a><br/> {}
</div>
</div>
"""
filepath = os.path.normpath( os.path.join(dir_path, filepath) )
# sort our scores dictionary to highest scoring and to 2 arrays
# this is a list of tupes of key values.
items = sorted(scores.items(), key = lambda kv:(kv[1], kv[0]))
items.reverse()
return html_entry.format(filepath, filepath, html_entry_scores_table(items ) )
def html_footer():
html_footer = """
</div>
</body>
</html>
"""
return html_footer
# open a file for writing
# for reference, for multi label we want to do
# : gs://calm-trees-123-vcm/flowers/images/5217892384_3edce91761_m.jpg,dandelion,tulip,rose
# from https://cloud.google.com/vision/automl/docs/prepare
start = time.time()
all_files = []
with open(args.output, 'wb') as writer:
if args.type == 'csv':
writer = csv.writer(writer)
else:
writer.write(html_header())
# recurse through our image directory and run inference on each image
for subdir, dirs, files in os.walk(args.imagedir):
for file in files:
#print os.path.join(subdir, file)
filepath = subdir + os.sep + file
if filepath.endswith(".jpg"):
all_files.append(filepath)
#do we shuffle our files?
if args.random == True:
random.shuffle(all_files)
#do we limit our file count so we can do a test run?
if args.limit is not 0:
all_files = all_files[:args.limit]
for filepath in all_files:
image = load_image(filepath, resize_to=(Width, Height))
if image != None:
labels = []
scores = {}
# prepend our prefix if we have it
if args.prefix:
labels.append( args.prefix + filepath )
else:
labels.append(filepath)
for model in models:
prediction = model.predict({'Image': image})
score = prediction['Scores']
scores.update(score)
label = prediction['Class Label']
labels.append(label)
#write all of our predictions out to our CSV
if args.type == 'csv':
writer.writerow(labels)
else:
# write HTML label version with file name for IMG tag, etc
if args.probabilities:
writer.write( html_entry_scores(filepath, scores) )
else:
writer.write( html_entry_label(filepath, labels) )
print("labeled " + filepath)
if args.type == 'html':
writer.write(html_footer())
end = time.time()
predictiontime = end - start
print("")
print("Completed Processing")
print("")
print( str( len(all_files) ) + " images processed in " + str(predictiontime) + " seconds")
print( str( len(all_files)/predictiontime ) + "images / second")
print("")