-
Notifications
You must be signed in to change notification settings - Fork 0
/
objectDetection.py
219 lines (169 loc) · 7.39 KB
/
objectDetection.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
import cv2 # state of the art computer vision algorithms library
import numpy as np # fundamental package for scientific computing
import matplotlib.pyplot as plt # 2D plotting library producing publication quality figures
import pyrealsense2 as rs # Intel RealSense cross-platform open-source API
print("Environment Ready")
def filtering(depth_frame):
decimation = rs.decimation_filter()
decimation.set_option(rs.option.filter_magnitude, 4)
spatial = rs.spatial_filter()
spatial.set_option(rs.option.filter_magnitude, 5)
spatial.set_option(rs.option.filter_smooth_alpha, 1)
spatial.set_option(rs.option.filter_smooth_delta, 50)
spatial.set_option(rs.option.holes_fill, 3)
hole_filling = rs.hole_filling_filter()
temporal = rs.temporal_filter()
depth_to_disparity = rs.disparity_transform(True)
disparity_to_depth = rs.disparity_transform(False)
frame = depth_frame
frame = decimation.process(frame)
frame = depth_to_disparity.process(frame)
frame = spatial.process(frame)
frame = temporal.process(frame)
frame = disparity_to_depth.process(frame)
depth_frame = hole_filling.process(frame)
return depth_frame
def colorSegementation(color, xmin, xmax, ymin, ymax):
# Estrealla red color
hsv_frame = cv2.cvtColor(color, cv2.COLOR_RGB2HSV)
mask1 = cv2.inRange(hsv_frame, (0,50,20), (5,255,255))
mask2 = cv2.inRange(hsv_frame, (175,50,20), (180,255,255))
## Merge the mask and crop the red regions
masked = cv2.bitwise_or(mask1, mask2)
red = cv2.bitwise_and(color, color, mask=masked)
indices = np.where(red != [0])
avg_x_ = None
avg_y_ = None
counter = 0
if len(indices[0]) and len(indices[1]):
avg_x_ = 0
avg_y_ = 0
counter = 0
for y in indices[0]:
if (y >= int(ymin) and y <= int(ymax)):
avg_y_ += y
counter += 1
try:
avg_y_ /= counter
except:
pass
counter = 0
for x in indices[1]:
if (x >= int(xmin) and x <= int(xmax)):
avg_x_ += x
counter += 1
try:
avg_x_ /= counter
except:
pass
#print("c:",counter)
if (avg_x_ >= int(xmin) and avg_x_ <= int(xmax)) and (avg_y_ >= int(ymin) and avg_y_ <= int(ymax)) and counter >= 3000: #counter is a hyperparameter
return "Estrella"
else:
return "Unknown"
# Setup:
pipe = rs.pipeline()
config = rs.config()
config.enable_stream(rs.stream.color, 424, 240, rs.format.rgb8, 30)
config.enable_stream(rs.stream.depth, 424, 240, rs.format.z16, 30)
config.enable_stream(rs.stream.infrared, 1)
config.enable_stream(rs.stream.infrared, 2)
profile = pipe.start(config)
# Skip 15 first frames to give the Auto-Exposure time to adjust
for x in range(15):
pipe.wait_for_frames()
# Store next frameset for later processing:
frameset = pipe.wait_for_frames()
color_frame = frameset.get_color_frame()
depth_frame = frameset.get_depth_frame()
depth_frame = filtering(depth_frame)
# Cleanup:
pipe.stop()
print("Frames Captured")
color = np.asanyarray(color_frame.get_data())
#plt.rcParams["axes.grid"] = False
#plt.rcParams['figure.figsize'] = [12, 6]
#plt.imshow(color)
#plt.show()
colorizer = rs.colorizer()
colorized_depth = np.asanyarray(colorizer.colorize(depth_frame).get_data())
#plt.imshow(colorized_depth)
# Create alignment primitive with color as its target stream:
align = rs.align(rs.stream.color)
frameset = align.process(frameset)
# Update color and depth frames:
aligned_depth_frame = frameset.get_depth_frame()
colorized_depth = np.asanyarray(colorizer.colorize(aligned_depth_frame).get_data())
# Show the two frames together:
#images = np.hstack((color, colorized_depth))
#plt.imshow(images)
# Standard OpenCV boilerplate for running the net:
height, width = color.shape[:2] #240, 424
expected = 300
aspect = width / height
resized_image = cv2.resize(color, (int(round(expected * aspect)), expected))
crop_start = int(round(expected * (aspect - 1) / 2))
crop_img = resized_image[0:expected, crop_start:crop_start+expected]
arg1 = "MobileNetSSD_deploy.prototxt.txt"
arg2 = "MobileNetSSD_deploy.caffemodel"
net = cv2.dnn.readNetFromCaffe(arg1, arg2)
inScaleFactor = 0.007843
meanVal = 127.53
classNames = ("background", "aeroplane", "bicycle", "bird", "boat",
"bottle", "bus", "car", "cat", "chair",
"cow", "diningtable", "dog", "horse",
"motorbike", "person", "pottedplant",
"sheep", "sofa", "train", "tvmonitor")
blob = cv2.dnn.blobFromImage(crop_img, inScaleFactor, (expected, expected), meanVal, False)
net.setInput(blob, "data")
detections = net.forward("detection_out")
results = []
for i in np.arange(0, detections.shape[2]):
idx = int(detections[0, 0, i, 1])
#print(classNames[idx])
label = detections[0,0,i,1]
conf = detections[0,0,i,2]
xmin = detections[0,0,i,3]
ymin = detections[0,0,i,4]
xmax = detections[0,0,i,5]
ymax = detections[0,0,i,6]
className = classNames[int(label)]
#cv2.rectangle(crop_img, (int(xmin * expected), int(ymin * expected)),
# (int(xmax * expected), int(ymax * expected)), (255, 255, 255), 2)
#cv2.putText(crop_img, className,
# (int(xmin * expected), int(ymin * expected) - 5),
# cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255))
#plt.imshow(crop_img)
#plt.show()
scale = height / expected
xmin_depth = int((xmin * expected + crop_start) * scale)
ymin_depth = int((ymin * expected) * scale)
xmax_depth = int((xmax * expected + crop_start) * scale)
ymax_depth = int((ymax * expected) * scale)
xmin_depth,ymin_depth,xmax_depth,ymax_depth
cv2.rectangle(colorized_depth, (xmin_depth, ymin_depth),
(xmax_depth, ymax_depth), (255, 255, 255), 2)
plt.imshow(colorized_depth)
# plt.show()
x_depth_center = 0.5 * (xmax_depth + xmin_depth)
y_depth_center = 0.5 * (ymax_depth + ymin_depth)
depth = np.asanyarray(aligned_depth_frame.get_data())
# Crop depth data:
depth = depth[xmin_depth:xmax_depth,ymin_depth:ymax_depth].astype(float)
# Get data scale from the device and convert to meters
#depth_scale = profile.get_device().first_depth_sensor().get_depth_scale()
#depth = depth * depth_scale
#dist,_,_,_ = cv2.mean(depth)
dist = aligned_depth_frame.get_distance(int(x_depth_center), int(y_depth_center))
#avg_x = 0.5 * (xmin * (width/expected) + xmax * (width/expected))
#avg_y = 0.5 * (ymin * (height/expected) + ymax * (height/expected))
#print(avg_x, avg_y)
depth_intrin = aligned_depth_frame.profile.as_video_stream_profile().intrinsics
depth = aligned_depth_frame.get_distance(int(x_depth_center), int(y_depth_center))
realx, realy, realz = rs.rs2_deproject_pixel_to_point(depth_intrin, [int(x_depth_center),int(y_depth_center)],depth)
objectType = colorSegementation(crop_img,int(xmin * expected), int(xmax * expected), int(ymin * expected), int(ymax * expected))
results.append((realx, realy, realz, objectType, className))
for item in results:
# if (item[3]=="bottle"):
print("Detected a {0} of type {4} at (x, y, z) : {1:.3}, {2:.3}, {3:.3}.".format(item[4], item[0], item[1], item[2], item[3]))
#The detections are in the list 'results' in the form of '(x, y, z, drinkType, className)'