-
Notifications
You must be signed in to change notification settings - Fork 1
/
final_model.py
283 lines (220 loc) · 8.61 KB
/
final_model.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
import tensorflow as tf
import numpy as np
import imutils
import os
import sys
from utility import *
from utils import label_map_util
from utils import visualization_utils as vis_util
from PIL import Image
import cv2
import math
import statistics
def bb_intersection_over_union(boxA, boxB):
# determine the (x, y)-coordinates of the intersection rectangle
xA = max(boxA["xmax"],boxB["xmax"])
yA = max(boxA["ymax"], boxB["ymax"])
xB = min(boxA["xmin"], boxB["xmin"])
yB = min(boxA["ymin"], boxB["ymin"])
# compute the area of intersection rectangle
interArea = max(0, xB - xA + 1) * max(0, yB - yA + 1)
# compute the area of both the prediction and ground-truth
# rectangles
boxAArea = (boxA["xmax"] - boxA["xmin"] + 1) * (boxA["ymin"] - boxA["ymax"] + 1)
boxBArea = (boxB["xmin"] - boxB["xmax"] + 1) * (boxB["ymin"] - boxB["ymax"] + 1)
# compute the intersection over union by taking the intersection
# area and dividing it by the sum of prediction + ground-truth
# areas - the interesection area
iou = interArea / float(boxAArea + boxBArea - interArea)
# return the intersection over union value
return iou
class ElectricMeterModel:
def __init__(self,model_name,path_to_ckpt,labels_path):
MODEL_NAME = model_name
PATH_TO_CKPT = os.path.join('data', MODEL_NAME + path_to_ckpt)
print(PATH_TO_CKPT)
# List of the strings that is used to add correct label for each box.
PATH_TO_LABELS = os.path.join('data', MODEL_NAME + labels_path)
print(PATH_TO_LABELS)
NUM_CLASSES = 40
# Loading label map
label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES,
use_display_name=True)
self.category_index = label_map_util.create_category_index(categories)
#create a tensorflow graph from graph definition from file
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
# l = [n.name for n in detection_graph.as_graph_def().node]
# print(l)
self.detection_graph = detection_graph
self.sess = tf.Session(graph=detection_graph)
#Used to convert opencv image to numpy array
def load_cvimage_into_numpy_array(self,image):
im_width = image.shape[1]
im_height = image.shape[0]
return image.reshape((im_height,im_width,3)).astype(np.uint8)
def load_image_into_numpy_array(self,image):
(im_width, im_height) = image.size
return np.array(image.getdata()).reshape((im_height, im_width, 3)).astype(np.uint8)
def detect_objects(self,image_np,sess,detection_graph):
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = self.detection_graph.get_tensor_by_name('image_tensor:0')
# Each box represents a part of the image where a particular object was detected.
boxes = self.detection_graph.get_tensor_by_name('detection_boxes:0')
# Each score represent how level of confidence for each of the objects.
# Score is shown on the result image, together with the class label.
scores = self.detection_graph.get_tensor_by_name('detection_scores:0')
classes = self.detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = self.detection_graph.get_tensor_by_name('num_detections:0')
# Actual detection.
(boxes, scores, classes, num_detections) = self.sess.run(
[boxes, scores, classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
# Visualization of the results of a detection.
rect_points, class_names, class_colors = draw_boxes_and_labels(
boxes=np.squeeze(boxes),
classes=np.squeeze(classes).astype(np.int32),
scores=np.squeeze(scores),
category_index=self.category_index,
min_score_thresh=.5
)
return dict(rect_points=rect_points, class_names=class_names, class_colors=class_colors)
def predict(self,cv_image):
exception = 1
image_np = self.load_cvimage_into_numpy_array(cv_image)
data = self.detect_objects(image_np,self.sess,self.detection_graph)
rec_points = data['rect_points']
class_names = data['class_names']
boxes = list(zip(rec_points, class_names))
#print(boxes)
#exctract labels for meter identificaton number
n_boxes = list(filter(lambda t: 'n' in t[1][0], boxes))
#extract labels for dials
dial_boxes = list(filter(lambda t: t not in n_boxes, boxes))
#print(dial_boxes)
#sort bounding rects by ymin
sorted_n_boxes = sorted(n_boxes, key=lambda tup: tup[0]['xmin'])
sorted_dial_boxes = sorted(dial_boxes, key=lambda tup: tup[0]['xmin'])
filtered_dial_boxes, exception = self.filter_iou(sorted_dial_boxes)
result_dict = {}
dial_readings = []
#go through and edit out x labels
dial_numbers = [el[1][0].split()[0][0:2] for el in filtered_dial_boxes]
dial_numbers.reverse()
final_reading = []
is_digital = False
for i, d in enumerate(dial_numbers):
if 'd' in d:
is_digital = True
if i==0:
final_reading.append(int(d[1]))
elif('x' in d):
if(int(final_reading[i-1]) >= 6):
if(int(d[1]) == 0):
final_reading.append(9)
else:
final_reading.append(int(d[1]) - 1)
else:
final_reading.append(int(d[1]))
else:
final_reading.append(int(d[1]))
final_reading.reverse()
meter_type = 'digital' if is_digital else 'analog'
for j, el in enumerate(filtered_dial_boxes):
split_prob = el[1][0].split()
prob = split_prob[1]
num = split_prob[0]
obj = {}
obj['number'] = final_reading[j]
obj['prob'] = prob
obj['dimen'] = el[0]
dial_readings.append(obj)
result_dict['dial_boxes'] = dial_readings
id_numbers = [el[1][0].split()[0][1] for el in sorted_n_boxes]
meter_id_numbers = []
for k, el in enumerate(sorted_n_boxes):
split_prob_n = el[1][0].split()
prob_n = split_prob_n[1]
num_n = split_prob_n[0]
obj_n = {}
obj_n['number'] = id_numbers[k]
obj_n['prob'] = prob_n
obj_n['dimen'] = el[0]
meter_id_numbers.append(obj_n)
result_dict['id_number'] = meter_id_numbers
#print(final_reading)
return meter_type, ''.join([str(f) for f in final_reading]), ''.join(id_numbers), result_dict, exception
# def filter_iou(self, sorted_boxes):
# exception = 1
# #handling intersections and double labels
# iou_lst = []
# el_removed = []
# if(len(sorted_boxes) > 5):
# exception = 2
# iou_lst = [bb_intersection_over_union(sorted_boxes[i][0],
# sorted_boxes[i+1][0]) for i in range(len(sorted_boxes)-1)]
# #find median iou
# med_iou = statistics.median(iou_lst)
# add_idxs = []
# for i in range(len(iou_lst) + 1):
# if i == len(iou_lst):
# if i not in add_idxs:
# el_removed.append(sorted_boxes[i])
# break
# elif(abs(iou_lst[i] - med_iou)>0.1):
# prob1 = float(sorted_boxes[i][1][0].split()[1][:2])
# prob2 = float(sorted_boxes[i+1][1][0].split()[1][:2])
# if(prob1 > prob2): #add the box that has the higher probability
# if(i not in add_idxs): #only add boxes that have not already been added
# el_removed.append(sorted_boxes[i])
# add_idxs.append(i)
# add_idxs.append(i+1)
# else:
# if(i+1 not in add_idxs):
# el_removed.append(sorted_boxes[i+1])
# add_idxs.append(i+1)
# add_idxs.append(i)
# else:
# if(i not in add_idxs):
# add_idxs.append(i)
# el_removed.append(sorted_boxes[i])
# else:
# if(len(sorted_boxes) < 5):
# exception = 3
# el_removed = sorted_boxes
# return el_removed, exception
def filter_iou(self, sorted_boxes):
def recursive_remove(boxes, med_iou):
if len(boxes) == 1:
return [boxes[0]]
elif(bb_intersection_over_union(boxes[0][0], #check if consecutive boxes overlap
boxes[1][0]) - med_iou > 0.1):
prob1 = float(boxes[0][1][0].split()[1][:2])
prob2 = float(boxes[1][1][0].split()[1][:2])
if(prob1 > prob2): #remove box with lower confidence
boxes.pop(1)
return recursive_remove(boxes, med_iou)
else:
boxes.pop(0)
return recursive_remove(boxes, med_iou)
else:
box_to_add = boxes[0]
boxes.pop(0)
return [box_to_add] + recursive_remove(boxes, med_iou)
med_iou = statistics.median([bb_intersection_over_union(sorted_boxes[i][0],
sorted_boxes[i+1][0]) for i in range(len(sorted_boxes)-1)])
extracted_boxes = recursive_remove(sorted_boxes, med_iou) #remove duplicate boxes
if(len(extracted_boxes) == 5):
if len(sorted_boxes) > 5:
exception = 2
else:
exception = 1
else:
exception = 3
return extracted_boxes, exception