質問編集履歴
3
リンクを追加しました
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -6,6 +6,10 @@
|
|
|
6
6
|
|
|
7
7
|
|
|
8
8
|
|
|
9
|
+
↓こちらからダウンロードしました。
|
|
10
|
+
|
|
11
|
+
[リンク内容](https://github.com/chandrikadeb7/Face-Mask-Detection)
|
|
12
|
+
|
|
9
13
|
|
|
10
14
|
|
|
11
15
|
### 該当のソースコード
|
2
Pythonと追加しました
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -14,6 +14,14 @@
|
|
|
14
14
|
|
|
15
15
|
```Python
|
|
16
16
|
|
|
17
|
+
# USAGE
|
|
18
|
+
|
|
19
|
+
# python detect_mask_video.py
|
|
20
|
+
|
|
21
|
+
|
|
22
|
+
|
|
23
|
+
# import the necessary packages
|
|
24
|
+
|
|
17
25
|
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
|
|
18
26
|
|
|
19
27
|
from tensorflow.keras.preprocessing.image import img_to_array
|
|
@@ -38,114 +46,166 @@
|
|
|
38
46
|
|
|
39
47
|
def detect_and_predict_mask(frame, faceNet, maskNet):
|
|
40
48
|
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
|
|
45
|
-
(
|
|
46
|
-
|
|
47
|
-
|
|
48
|
-
|
|
49
|
-
|
|
50
|
-
|
|
51
|
-
|
|
52
|
-
|
|
53
|
-
|
|
54
|
-
|
|
55
|
-
faces
|
|
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
|
-
i
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
f
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
96
|
-
|
|
97
|
-
|
|
98
|
-
|
|
99
|
-
|
|
100
|
-
|
|
101
|
-
|
|
102
|
-
|
|
103
|
-
|
|
104
|
-
|
|
105
|
-
r
|
|
106
|
-
|
|
107
|
-
|
|
49
|
+
# grab the dimensions of the frame and then construct a blob
|
|
50
|
+
|
|
51
|
+
# from it
|
|
52
|
+
|
|
53
|
+
(h, w) = frame.shape[:2]
|
|
54
|
+
|
|
55
|
+
blob = cv2.dnn.blobFromImage(frame, 1.0, (300, 300),
|
|
56
|
+
|
|
57
|
+
(104.0, 177.0, 123.0))
|
|
58
|
+
|
|
59
|
+
|
|
60
|
+
|
|
61
|
+
# pass the blob through the network and obtain the face detections
|
|
62
|
+
|
|
63
|
+
faceNet.setInput(blob)
|
|
64
|
+
|
|
65
|
+
detections = faceNet.forward()
|
|
66
|
+
|
|
67
|
+
|
|
68
|
+
|
|
69
|
+
# initialize our list of faces, their corresponding locations,
|
|
70
|
+
|
|
71
|
+
# and the list of predictions from our face mask network
|
|
72
|
+
|
|
73
|
+
faces = []
|
|
74
|
+
|
|
75
|
+
locs = []
|
|
76
|
+
|
|
77
|
+
preds = []
|
|
78
|
+
|
|
79
|
+
|
|
80
|
+
|
|
81
|
+
# loop over the detections
|
|
82
|
+
|
|
83
|
+
for i in range(0, detections.shape[2]):
|
|
84
|
+
|
|
85
|
+
# extract the confidence (i.e., probability) associated with
|
|
86
|
+
|
|
87
|
+
# the detection
|
|
88
|
+
|
|
89
|
+
confidence = detections[0, 0, i, 2]
|
|
90
|
+
|
|
91
|
+
|
|
92
|
+
|
|
93
|
+
# filter out weak detections by ensuring the confidence is
|
|
94
|
+
|
|
95
|
+
# greater than the minimum confidence
|
|
96
|
+
|
|
97
|
+
if confidence > args["confidence"]:
|
|
98
|
+
|
|
99
|
+
# compute the (x, y)-coordinates of the bounding box for
|
|
100
|
+
|
|
101
|
+
# the object
|
|
102
|
+
|
|
103
|
+
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
|
|
104
|
+
|
|
105
|
+
(startX, startY, endX, endY) = box.astype("int")
|
|
106
|
+
|
|
107
|
+
|
|
108
|
+
|
|
109
|
+
# ensure the bounding boxes fall within the dimensions of
|
|
110
|
+
|
|
111
|
+
# the frame
|
|
112
|
+
|
|
113
|
+
(startX, startY) = (max(0, startX), max(0, startY))
|
|
114
|
+
|
|
115
|
+
(endX, endY) = (min(w - 1, endX), min(h - 1, endY))
|
|
116
|
+
|
|
117
|
+
|
|
118
|
+
|
|
119
|
+
# extract the face ROI, convert it from BGR to RGB channel
|
|
120
|
+
|
|
121
|
+
# ordering, resize it to 224x224, and preprocess it
|
|
122
|
+
|
|
123
|
+
face = frame[startY:endY, startX:endX]
|
|
124
|
+
|
|
125
|
+
if face.any():
|
|
126
|
+
|
|
127
|
+
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
|
|
128
|
+
|
|
129
|
+
face = cv2.resize(face, (224, 224))
|
|
130
|
+
|
|
131
|
+
face = img_to_array(face)
|
|
132
|
+
|
|
133
|
+
face = preprocess_input(face)
|
|
134
|
+
|
|
135
|
+
|
|
136
|
+
|
|
137
|
+
# only make a predictions if at least one face was detected
|
|
138
|
+
|
|
139
|
+
if len(faces) > 0:
|
|
140
|
+
|
|
141
|
+
# for faster inference we'll make batch predictions on *all*
|
|
142
|
+
|
|
143
|
+
# faces at the same time rather than one-by-one predictions
|
|
144
|
+
|
|
145
|
+
# in the above `for` loop
|
|
146
|
+
|
|
147
|
+
faces = np.array(faces, dtype="float32")
|
|
148
|
+
|
|
149
|
+
preds = maskNet.predict(faces, batch_size=32)
|
|
150
|
+
|
|
151
|
+
|
|
152
|
+
|
|
153
|
+
# return a 2-tuple of the face locations and their corresponding
|
|
154
|
+
|
|
155
|
+
# locations
|
|
156
|
+
|
|
157
|
+
return (locs, preds)
|
|
158
|
+
|
|
159
|
+
|
|
160
|
+
|
|
161
|
+
# construct the argument parser and parse the arguments
|
|
108
162
|
|
|
109
163
|
ap = argparse.ArgumentParser()
|
|
110
164
|
|
|
111
165
|
ap.add_argument("-f", "--face", type=str,
|
|
112
166
|
|
|
113
|
-
default="face_detector",
|
|
167
|
+
default="face_detector",
|
|
114
|
-
|
|
168
|
+
|
|
115
|
-
help="path to face detector model directory")
|
|
169
|
+
help="path to face detector model directory")
|
|
116
170
|
|
|
117
171
|
ap.add_argument("-m", "--model", type=str,
|
|
118
172
|
|
|
119
|
-
default="mask_detector.model",
|
|
173
|
+
default="mask_detector.model",
|
|
120
|
-
|
|
174
|
+
|
|
121
|
-
help="path to trained face mask detector model")
|
|
175
|
+
help="path to trained face mask detector model")
|
|
122
176
|
|
|
123
177
|
ap.add_argument("-c", "--confidence", type=float, default=0.5,
|
|
124
178
|
|
|
125
|
-
help="minimum probability to filter weak detections")
|
|
179
|
+
help="minimum probability to filter weak detections")
|
|
126
180
|
|
|
127
181
|
args = vars(ap.parse_args())
|
|
128
182
|
|
|
129
183
|
|
|
130
184
|
|
|
185
|
+
# load our serialized face detector model from disk
|
|
186
|
+
|
|
131
187
|
print("[INFO] loading face detector model...")
|
|
132
188
|
|
|
133
189
|
prototxtPath = os.path.sep.join([args["face"], "deploy.prototxt"])
|
|
134
190
|
|
|
135
191
|
weightsPath = os.path.sep.join([args["face"],
|
|
136
192
|
|
|
137
|
-
"res10_300x300_ssd_iter_140000.caffemodel"])
|
|
193
|
+
"res10_300x300_ssd_iter_140000.caffemodel"])
|
|
138
194
|
|
|
139
195
|
faceNet = cv2.dnn.readNet(prototxtPath, weightsPath)
|
|
140
196
|
|
|
141
197
|
|
|
142
198
|
|
|
199
|
+
# load the face mask detector model from disk
|
|
200
|
+
|
|
143
201
|
print("[INFO] loading face mask detector model...")
|
|
144
202
|
|
|
145
203
|
maskNet = load_model(args["model"])
|
|
146
204
|
|
|
147
205
|
|
|
148
206
|
|
|
207
|
+
# initialize the video stream and allow the camera sensor to warm up
|
|
208
|
+
|
|
149
209
|
print("[INFO] starting video stream...")
|
|
150
210
|
|
|
151
211
|
vs = VideoStream(src=0).start()
|
|
@@ -154,58 +214,92 @@
|
|
|
154
214
|
|
|
155
215
|
|
|
156
216
|
|
|
217
|
+
# loop over the frames from the video stream
|
|
218
|
+
|
|
157
219
|
while True:
|
|
158
220
|
|
|
159
|
-
frame
|
|
160
|
-
|
|
161
|
-
|
|
162
|
-
|
|
163
|
-
|
|
164
|
-
|
|
165
|
-
|
|
166
|
-
|
|
167
|
-
|
|
168
|
-
|
|
169
|
-
f
|
|
170
|
-
|
|
171
|
-
|
|
172
|
-
|
|
173
|
-
(
|
|
174
|
-
|
|
175
|
-
|
|
176
|
-
|
|
177
|
-
l
|
|
178
|
-
|
|
179
|
-
|
|
180
|
-
|
|
181
|
-
|
|
182
|
-
|
|
183
|
-
|
|
184
|
-
|
|
185
|
-
|
|
186
|
-
|
|
187
|
-
|
|
188
|
-
|
|
189
|
-
|
|
190
|
-
|
|
191
|
-
|
|
192
|
-
|
|
193
|
-
|
|
194
|
-
|
|
195
|
-
|
|
196
|
-
|
|
197
|
-
|
|
198
|
-
|
|
199
|
-
|
|
200
|
-
|
|
201
|
-
i
|
|
202
|
-
|
|
203
|
-
break
|
|
204
|
-
|
|
205
|
-
|
|
221
|
+
# grab the frame from the threaded video stream and resize it
|
|
222
|
+
|
|
223
|
+
# to have a maximum width of 400 pixels
|
|
224
|
+
|
|
225
|
+
frame = vs.read()
|
|
226
|
+
|
|
227
|
+
frame = imutils.resize(frame, width=400)
|
|
228
|
+
|
|
229
|
+
|
|
230
|
+
|
|
231
|
+
# detect faces in the frame and determine if they are wearing a
|
|
232
|
+
|
|
233
|
+
# face mask or not
|
|
234
|
+
|
|
235
|
+
(locs, preds) = detect_and_predict_mask(frame, faceNet, maskNet)
|
|
236
|
+
|
|
237
|
+
|
|
238
|
+
|
|
239
|
+
# loop over the detected face locations and their corresponding
|
|
240
|
+
|
|
241
|
+
# locations
|
|
242
|
+
|
|
243
|
+
for (box, pred) in zip(locs, preds):
|
|
244
|
+
|
|
245
|
+
# unpack the bounding box and predictions
|
|
246
|
+
|
|
247
|
+
(startX, startY, endX, endY) = box
|
|
248
|
+
|
|
249
|
+
(mask, withoutMask) = pred
|
|
250
|
+
|
|
251
|
+
|
|
252
|
+
|
|
253
|
+
# determine the class label and color we'll use to draw
|
|
254
|
+
|
|
255
|
+
# the bounding box and text
|
|
256
|
+
|
|
257
|
+
label = "Mask" if mask > withoutMask else "No Mask"
|
|
258
|
+
|
|
259
|
+
color = (0, 255, 0) if label == "Mask" else (0, 0, 255)
|
|
260
|
+
|
|
261
|
+
|
|
262
|
+
|
|
263
|
+
# include the probability in the label
|
|
264
|
+
|
|
265
|
+
label = "{}: {:.2f}%".format(label, max(mask, withoutMask) * 100)
|
|
266
|
+
|
|
267
|
+
|
|
268
|
+
|
|
269
|
+
# display the label and bounding box rectangle on the output
|
|
270
|
+
|
|
271
|
+
# frame
|
|
272
|
+
|
|
273
|
+
cv2.putText(frame, label, (startX, startY - 10),
|
|
274
|
+
|
|
275
|
+
cv2.FONT_HERSHEY_SIMPLEX, 0.45, color, 2)
|
|
276
|
+
|
|
277
|
+
cv2.rectangle(frame, (startX, startY), (endX, endY), color, 2)
|
|
278
|
+
|
|
279
|
+
|
|
280
|
+
|
|
281
|
+
# show the output frame
|
|
282
|
+
|
|
283
|
+
cv2.imshow("Frame", frame)
|
|
284
|
+
|
|
285
|
+
key = cv2.waitKey(1) & 0xFF
|
|
286
|
+
|
|
287
|
+
|
|
288
|
+
|
|
289
|
+
# if the `q` key was pressed, break from the loop
|
|
290
|
+
|
|
291
|
+
if key == ord("q"):
|
|
292
|
+
|
|
293
|
+
break
|
|
294
|
+
|
|
295
|
+
|
|
296
|
+
|
|
297
|
+
# do a bit of cleanup
|
|
206
298
|
|
|
207
299
|
cv2.destroyAllWindows()
|
|
208
300
|
|
|
209
301
|
vs.stop()
|
|
210
302
|
|
|
303
|
+
|
|
304
|
+
|
|
211
305
|
```
|
1
Pythonと追加しました
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -12,6 +12,8 @@
|
|
|
12
12
|
|
|
13
13
|
|
|
14
14
|
|
|
15
|
+
```Python
|
|
16
|
+
|
|
15
17
|
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
|
|
16
18
|
|
|
17
19
|
from tensorflow.keras.preprocessing.image import img_to_array
|
|
@@ -205,3 +207,5 @@
|
|
|
205
207
|
cv2.destroyAllWindows()
|
|
206
208
|
|
|
207
209
|
vs.stop()
|
|
210
|
+
|
|
211
|
+
```
|