質問編集履歴
2
リンクを追加しました
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -8,6 +8,12 @@
|
|
|
8
8
|
|
|
9
9
|
|
|
10
10
|
|
|
11
|
+
githubよりダウンロードしました
|
|
12
|
+
|
|
13
|
+
[リンク内容](https://github.com/chandrikadeb7/Face-Mask-Detection)
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
|
|
11
17
|
|
|
12
18
|
|
|
13
19
|
```Python
|
1
Pythonと追加しました
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -10,7 +10,15 @@
|
|
|
10
10
|
|
|
11
11
|
|
|
12
12
|
|
|
13
|
+
```Python
|
|
14
|
+
|
|
13
|
-
#
|
|
15
|
+
# USAGE
|
|
16
|
+
|
|
17
|
+
# python detect_mask_video.py
|
|
18
|
+
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
# import the necessary packages
|
|
14
22
|
|
|
15
23
|
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
|
|
16
24
|
|
|
@@ -36,6 +44,10 @@
|
|
|
36
44
|
|
|
37
45
|
def detect_and_predict_mask(frame, faceNet, maskNet):
|
|
38
46
|
|
|
47
|
+
# grab the dimensions of the frame and then construct a blob
|
|
48
|
+
|
|
49
|
+
# from it
|
|
50
|
+
|
|
39
51
|
(h, w) = frame.shape[:2]
|
|
40
52
|
|
|
41
53
|
blob = cv2.dnn.blobFromImage(frame, 1.0, (300, 300),
|
|
@@ -44,12 +56,18 @@
|
|
|
44
56
|
|
|
45
57
|
|
|
46
58
|
|
|
59
|
+
# pass the blob through the network and obtain the face detections
|
|
60
|
+
|
|
47
61
|
faceNet.setInput(blob)
|
|
48
62
|
|
|
49
63
|
detections = faceNet.forward()
|
|
50
64
|
|
|
51
65
|
|
|
52
66
|
|
|
67
|
+
# initialize our list of faces, their corresponding locations,
|
|
68
|
+
|
|
69
|
+
# and the list of predictions from our face mask network
|
|
70
|
+
|
|
53
71
|
faces = []
|
|
54
72
|
|
|
55
73
|
locs = []
|
|
@@ -58,26 +76,48 @@
|
|
|
58
76
|
|
|
59
77
|
|
|
60
78
|
|
|
79
|
+
# loop over the detections
|
|
80
|
+
|
|
61
81
|
for i in range(0, detections.shape[2]):
|
|
62
82
|
|
|
83
|
+
# extract the confidence (i.e., probability) associated with
|
|
84
|
+
|
|
85
|
+
# the detection
|
|
86
|
+
|
|
63
87
|
confidence = detections[0, 0, i, 2]
|
|
64
88
|
|
|
65
89
|
|
|
66
90
|
|
|
91
|
+
# filter out weak detections by ensuring the confidence is
|
|
92
|
+
|
|
93
|
+
# greater than the minimum confidence
|
|
94
|
+
|
|
67
95
|
if confidence > args["confidence"]:
|
|
68
96
|
|
|
97
|
+
# compute the (x, y)-coordinates of the bounding box for
|
|
98
|
+
|
|
99
|
+
# the object
|
|
100
|
+
|
|
69
101
|
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
|
|
70
102
|
|
|
71
103
|
(startX, startY, endX, endY) = box.astype("int")
|
|
72
104
|
|
|
73
105
|
|
|
74
106
|
|
|
107
|
+
# ensure the bounding boxes fall within the dimensions of
|
|
108
|
+
|
|
109
|
+
# the frame
|
|
110
|
+
|
|
75
111
|
(startX, startY) = (max(0, startX), max(0, startY))
|
|
76
112
|
|
|
77
113
|
(endX, endY) = (min(w - 1, endX), min(h - 1, endY))
|
|
78
114
|
|
|
79
115
|
|
|
80
116
|
|
|
117
|
+
# extract the face ROI, convert it from BGR to RGB channel
|
|
118
|
+
|
|
119
|
+
# ordering, resize it to 224x224, and preprocess it
|
|
120
|
+
|
|
81
121
|
face = frame[startY:endY, startX:endX]
|
|
82
122
|
|
|
83
123
|
if face.any():
|
|
@@ -92,18 +132,32 @@
|
|
|
92
132
|
|
|
93
133
|
|
|
94
134
|
|
|
135
|
+
# only make a predictions if at least one face was detected
|
|
136
|
+
|
|
95
137
|
if len(faces) > 0:
|
|
96
138
|
|
|
139
|
+
# for faster inference we'll make batch predictions on *all*
|
|
140
|
+
|
|
141
|
+
# faces at the same time rather than one-by-one predictions
|
|
142
|
+
|
|
143
|
+
# in the above `for` loop
|
|
144
|
+
|
|
97
145
|
faces = np.array(faces, dtype="float32")
|
|
98
146
|
|
|
99
147
|
preds = maskNet.predict(faces, batch_size=32)
|
|
100
148
|
|
|
101
149
|
|
|
102
150
|
|
|
151
|
+
# return a 2-tuple of the face locations and their corresponding
|
|
152
|
+
|
|
153
|
+
# locations
|
|
154
|
+
|
|
103
155
|
return (locs, preds)
|
|
104
156
|
|
|
105
157
|
|
|
106
158
|
|
|
159
|
+
# construct the argument parser and parse the arguments
|
|
160
|
+
|
|
107
161
|
ap = argparse.ArgumentParser()
|
|
108
162
|
|
|
109
163
|
ap.add_argument("-f", "--face", type=str,
|
|
@@ -126,6 +180,8 @@
|
|
|
126
180
|
|
|
127
181
|
|
|
128
182
|
|
|
183
|
+
# load our serialized face detector model from disk
|
|
184
|
+
|
|
129
185
|
print("[INFO] loading face detector model...")
|
|
130
186
|
|
|
131
187
|
prototxtPath = os.path.sep.join([args["face"], "deploy.prototxt"])
|
|
@@ -138,12 +194,16 @@
|
|
|
138
194
|
|
|
139
195
|
|
|
140
196
|
|
|
197
|
+
# load the face mask detector model from disk
|
|
198
|
+
|
|
141
199
|
print("[INFO] loading face mask detector model...")
|
|
142
200
|
|
|
143
201
|
maskNet = load_model(args["model"])
|
|
144
202
|
|
|
145
203
|
|
|
146
204
|
|
|
205
|
+
# initialize the video stream and allow the camera sensor to warm up
|
|
206
|
+
|
|
147
207
|
print("[INFO] starting video stream...")
|
|
148
208
|
|
|
149
209
|
vs = VideoStream(src=0).start()
|
|
@@ -152,36 +212,62 @@
|
|
|
152
212
|
|
|
153
213
|
|
|
154
214
|
|
|
215
|
+
# loop over the frames from the video stream
|
|
216
|
+
|
|
155
217
|
while True:
|
|
156
218
|
|
|
219
|
+
# grab the frame from the threaded video stream and resize it
|
|
220
|
+
|
|
221
|
+
# to have a maximum width of 400 pixels
|
|
222
|
+
|
|
157
223
|
frame = vs.read()
|
|
158
224
|
|
|
159
225
|
frame = imutils.resize(frame, width=400)
|
|
160
226
|
|
|
161
227
|
|
|
162
228
|
|
|
229
|
+
# detect faces in the frame and determine if they are wearing a
|
|
230
|
+
|
|
231
|
+
# face mask or not
|
|
232
|
+
|
|
163
233
|
(locs, preds) = detect_and_predict_mask(frame, faceNet, maskNet)
|
|
164
234
|
|
|
165
235
|
|
|
166
236
|
|
|
237
|
+
# loop over the detected face locations and their corresponding
|
|
238
|
+
|
|
239
|
+
# locations
|
|
240
|
+
|
|
167
241
|
for (box, pred) in zip(locs, preds):
|
|
168
242
|
|
|
243
|
+
# unpack the bounding box and predictions
|
|
244
|
+
|
|
169
245
|
(startX, startY, endX, endY) = box
|
|
170
246
|
|
|
171
247
|
(mask, withoutMask) = pred
|
|
172
248
|
|
|
173
249
|
|
|
174
250
|
|
|
251
|
+
# determine the class label and color we'll use to draw
|
|
252
|
+
|
|
253
|
+
# the bounding box and text
|
|
254
|
+
|
|
175
255
|
label = "Mask" if mask > withoutMask else "No Mask"
|
|
176
256
|
|
|
177
257
|
color = (0, 255, 0) if label == "Mask" else (0, 0, 255)
|
|
178
258
|
|
|
179
259
|
|
|
180
260
|
|
|
261
|
+
# include the probability in the label
|
|
262
|
+
|
|
181
263
|
label = "{}: {:.2f}%".format(label, max(mask, withoutMask) * 100)
|
|
182
264
|
|
|
183
265
|
|
|
184
266
|
|
|
267
|
+
# display the label and bounding box rectangle on the output
|
|
268
|
+
|
|
269
|
+
# frame
|
|
270
|
+
|
|
185
271
|
cv2.putText(frame, label, (startX, startY - 10),
|
|
186
272
|
|
|
187
273
|
cv2.FONT_HERSHEY_SIMPLEX, 0.45, color, 2)
|
|
@@ -190,18 +276,28 @@
|
|
|
190
276
|
|
|
191
277
|
|
|
192
278
|
|
|
279
|
+
# show the output frame
|
|
280
|
+
|
|
193
281
|
cv2.imshow("Frame", frame)
|
|
194
282
|
|
|
195
283
|
key = cv2.waitKey(1) & 0xFF
|
|
196
284
|
|
|
197
285
|
|
|
198
286
|
|
|
287
|
+
# if the `q` key was pressed, break from the loop
|
|
288
|
+
|
|
199
289
|
if key == ord("q"):
|
|
200
290
|
|
|
201
291
|
break
|
|
202
292
|
|
|
203
293
|
|
|
204
294
|
|
|
295
|
+
# do a bit of cleanup
|
|
296
|
+
|
|
205
297
|
cv2.destroyAllWindows()
|
|
206
298
|
|
|
207
299
|
vs.stop()
|
|
300
|
+
|
|
301
|
+
|
|
302
|
+
|
|
303
|
+
```
|