/
classify.py
executable file
·179 lines (128 loc) · 5.54 KB
/
classify.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
#!/usr/bin/python
# Martin Kersner, m.kersner@gmail.com
# 2015/11/26
import sys
import os
import csv
import label_init as li
import tools as tl
from ProgressBar import *
from ConfusionMatrix import *
import numpy as np
tl.load_caffe()
import caffe
import matplotlib.pyplot as plt
from matplotlib.colors import LogNorm
from collections import Counter
def main():
tl.check_arguments(sys.argv, 1, "You have to specify settings file!\n./classify.py settings_file")
settings_filename = sys.argv[1]
settings = tl.load_settings(settings_filename)
caffe_prototxt = settings["caffe_prototxt"]
caffe_model = settings["caffe_model"]
test_list = settings["test_list"]
caffe.set_mode_gpu()
net = caffe.Net(caffe_prototxt, caffe_model, caffe.TEST)
transformer = create_transformer(net)
img_test_names, img_test_labels = tl.read_img_names_from_csv(test_list,
skip_header=False,
delimiter=',')
submission_file, confusion_matrix, accuracy = classify_images(net,
transformer,
img_test_names,
img_test_labels,
settings)
if (settings["print_csv"]):
print "Results written to " + submission_file
print "Accuracy: " + str(accuracy)
# Confusion matrix
if (settings["conf_matrix"]):
plt.figure(figsize=(2,2), dpi=200)
plt.clf()
plt.imshow(confusion_matrix, norm=LogNorm())
#plt.colorbar()
plt.savefig(settings["conf_matrix_path"])
def create_transformer(net):
transformer = caffe.io.Transformer({"data": net.blobs["data"].data.shape})
transformer.set_transpose("data", (2,0,1))
transformer.set_raw_scale("data", 255)
transformer.set_channel_swap("data", (2,1,0))
return transformer
def classify_images(net, transformer, img_test_names, img_test_labels, settings):
pb = ProgressBar(len(img_test_names))
cm = ConfusionMatrix(li.labels)
pred_hit = np.zeros(len(img_test_names))
class_preds = []
for i, (id_, true_class) in enumerate(zip(img_test_names, img_test_labels)):
pb.print_progress()
if (settings["bulk"]):
settings["test_dir"] = settings["bulk_dir"]
pred_class = classify_from_bulk(net, transformer, id_, true_class, settings)
else:
pred_class = classify_image(net, transformer, id_, settings)[0]
class_preds.append(pred_class)
if (pred_class == true_class):
pred_hit[i] = 1
cm.actualize(true_class, pred_class)
submission_file = print_submission(img_test_names, class_preds, settings)
accuracy = compute_accuracy(pred_hit)
return submission_file, cm.get_confusion_matrix(), accuracy
def print_submission(img_test_names, class_preds, settings):
submission_file = None
if (settings["print_csv"]):
submission_file = print_csv(img_test_names, class_preds, settings)
elif (settings["print_term"]):
print_term(img_test_names, class_preds, settings)
return submission_file
def print_csv(img_ids, class_preds, settings):
field_names = settings["field_names"]
fn_id = field_names[0]
fn_class = field_names[1]
submission_file = "submission-" + tl.current_time() + ".csv"
with open(submission_file, 'wb') as csvfile:
writer = csv.DictWriter(csvfile, delimiter=',', fieldnames=field_names)
writer.writeheader()
for id_, pred in zip(img_ids, class_preds):
writer.writerow({fn_id: id_, fn_class: pred})
return submission_file
def print_term(img_ids, class_preds, settings):
fn_id = settings["field_names"][0]
fn_class = settings["field_names"][1]
print fn_id + "," + fn_class
for id_, pred in zip(img_ids, class_preds):
print id_ + "," + pred
def compute_accuracy(pred_hit):
return (1.0*np.sum(pred_hit))/len(pred_hit)
def classify_from_bulk(net, transformer, id_, true_class, settings):
bulk_suffixes = settings["bulk_suffixes"]
bulk_size = settings["bulk_size"]
pred_classes = []
for suffix in bulk_suffixes:
img_name = create_bulk_name([id_, true_class, suffix], '-')
pred_classes.extend(classify_image(net, transformer, img_name, settings)[0:bulk_size])
class_counter = Counter(pred_classes)
return most_frequent(class_counter)
def create_bulk_name(names, delimiter):
bulk_name = ""
last_name = names.pop()
for name in names:
bulk_name += name + delimiter
bulk_name += last_name
return bulk_name
def most_frequent(class_counter):
return class_counter.most_common(1)[0][0]
def classify_image(net, transformer, id_, settings):
img_width = settings["img_width"]
img_height = settings["img_height"]
test_dir = settings["test_dir"]
img_format = settings["img_format"]
img_path = os.path.join(test_dir, id_ + "." + img_format)
net.blobs["data"].reshape(1, 3, img_width, img_height) # does it have to be executed for each image?
net.blobs["data"].data[...] = transformer.preprocess("data", caffe.io.load_image(img_path))
out = net.forward()
# return all predicted classes sorted according to prob
sort_idxs = out["prob"][0].argsort()[::-1]
classes = [li.inv_labels[cls] for cls in sort_idxs]
return classes
if __name__ == "__main__":
main()