Skip to content

Commit 151f32b

Browse files
PlayerX5PlayerX5
committed
Face Mask Detection
0 parents  commit 151f32b

11 files changed

+2457
-0
lines changed

‎Complex_One.py‎

Lines changed: 189 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,189 @@
1+
# Python Face_Mask_Detection 1.py
2+
3+
# Import the necessary packages
4+
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
5+
from tensorflow.keras.preprocessing.image import img_to_array
6+
from tensorflow.keras.models import load_model
7+
from imutils.video import VideoStream
8+
import numpy as np
9+
import argparse
10+
import imutils
11+
import time
12+
import cv2
13+
import os
14+
import serial
15+
import PySimpleGUI as sg
16+
17+
sg.theme("GreenTan")
18+
19+
layout_b = [
20+
[sg.Frame(
21+
layout=[
22+
[sg.Text("Choose any one of the below actions")],
23+
],
24+
title="", relief=sg.RELIEF_GROOVE, )
25+
],
26+
[sg.Button("Camera", auto_size_button=True, tooltip="Press it!!"),
27+
sg.Button("Photo", auto_size_button=True, tooltip="Hey"),
28+
sg.Graph((125, 50), (0, 0), (125, 50), k='-GRAPH-')],
29+
]
30+
31+
layout = [
32+
[sg.T('The Face Mask Detection', font='_ 18', justification='c', expand_x=True, tooltip="Title")],
33+
[[layout_b]]
34+
]
35+
36+
window = sg.Window("The PySimpleGUI", layout, finalize=True, keep_on_top=True)
37+
window['-GRAPH-'].draw_image(data=sg.EMOJI_BASE64_DREAMING, location=(0, 50))
38+
39+
while True:
40+
event, values = window.read()
41+
42+
if event == sg.WIN_CLOSED:
43+
ch = sg.popup_yes_no("Are you sure you want to exit?", title="YesNo")
44+
if ch == "Yes":
45+
break
46+
47+
elif event == "Camera":
48+
try:
49+
def detect_and_predict_mask(frame, faceNet, maskNet):
50+
# Grab the dimensions of the frame and then construct a blob from it
51+
(h, w) = frame.shape[:2]
52+
blob = cv2.dnn.blobFromImage(frame, 1.0, (500, 500), (104.0, 177.0, 123.0))
53+
54+
# Pass the blob through the network and obtain the face detections
55+
faceNet.setInput(blob)
56+
detections = faceNet.forward()
57+
58+
# Initialize our list of faces, their corresponding locations and the list of predictions from our
59+
# face mask network
60+
faces = []
61+
locs = []
62+
preds = []
63+
64+
# Loop over the detections
65+
for i in range(0, detections.shape[2]):
66+
# Extract the confidence (i.e., probability) associated with the detection
67+
confidence = detections[0, 0, i, 2]
68+
69+
# Filter out weak detections by ensuring the confidence is greater than the minimum confidence
70+
if confidence > args["confidence"]:
71+
# Compute the (x, y)-coordinates of the bounding box for the object
72+
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
73+
(startX, startY, endX, endY) = box.astype("int")
74+
75+
# Ensure the bounding boxes fall within the dimensions of the frame
76+
(startX, startY) = (max(0, startX), max(0, startY))
77+
(endX, endY) = (min(w - 1, endX), min(h - 1, endY))
78+
79+
# Extract the face ROI(Region of Interest), convert it from BGR to RGB channel ordering,
80+
# resize it to 224x224, and preprocess it
81+
face = frame[startY:endY, startX:endX]
82+
if face.any():
83+
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
84+
face = cv2.resize(face, (224, 224))
85+
face = img_to_array(face)
86+
face = preprocess_input(face)
87+
88+
# Add the face and bounding boxes to their respective lists
89+
faces.append(face)
90+
locs.append((startX, startY, endX, endY))
91+
92+
# Only make a predictions if at least one face was detected
93+
if len(faces) == 1:
94+
# For faster inference we'll make batch predictions on *all* faces at the same time
95+
# rather than one-by-one predictions in the above `for` loop
96+
faces = np.array(faces, dtype="float32")
97+
preds = maskNet.predict(faces, batch_size=32)
98+
99+
# Return a 2-tuple(Ordered Pair, For holding two values together of any data type) of the face locations
100+
# and their corresponding locations
101+
return locs, preds
102+
103+
104+
# Construct the argument parser(To read the file containing the arguments) and parse the arguments
105+
ap = argparse.ArgumentParser()
106+
ap.add_argument("-f", "--face", type=str, default="face_detector",
107+
help="path to face detector model directory")
108+
ap.add_argument("-m", "--model", type=str, default="mask_detector.model",
109+
help="path to trained face mask detector model")
110+
ap.add_argument("-c", "--confidence", type=float, default=0.5,
111+
help="minimum probability to filter weak detections")
112+
args = vars(ap.parse_args())
113+
114+
# Load our serialized face detector model from disk os.path.sep.join is a method used to join components
115+
# of a file with the appropriate seperator for the current os
116+
print("[INFO] loading face detector model...")
117+
prototxtPath = os.path.sep.join(
118+
[args["face"], "deploy.prototxt"]) # .prototxt defines the architecture and parameters
119+
weightsPath = os.path.sep.join([args["face"],
120+
"res10_300x300_ssd_iter_140000.caffemodel"]) # .caffemodel contains the learned
121+
# weights of the model
122+
faceNet = cv2.dnn.readNet(prototxtPath,
123+
weightsPath) # cv2.dnn.readNet is a function from OpenCV's dnn module used
124+
# to read learning network models
125+
126+
# Load the face mask detector model from disk
127+
print("[INFO] loading face mask detector model...")
128+
maskNet = load_model("model.h5")
129+
130+
# Initialize the video stream and allow the camera sensor to warm up
131+
print("[INFO] starting video stream...")
132+
vs = VideoStream(src=0).start()
133+
time.sleep(2.0)
134+
135+
# Loop over the frames from the video stream
136+
while True:
137+
# Grab the frame from the threaded video stream and resize it to have a maximum width of 500 pixels (
138+
# Chose 400 pixels as a balance between reducing computational load while still maintaining enough
139+
# detail for the application's needs)
140+
frame = vs.read()
141+
frame = imutils.resize(frame, width=500)
142+
143+
# Detect faces in the frame and determine if they are wearing a face mask or not
144+
(locs, preds) = detect_and_predict_mask(frame, faceNet, maskNet)
145+
146+
# Loop over the detected face locations and their corresponding locations
147+
for (box, pred) in zip(locs, preds):
148+
# Unpack the bounding box and predictions
149+
(startX, startY, endX, endY) = box
150+
(mask, withoutMask) = pred
151+
152+
# Determine the class label and color we will use to draw the bounding box and text
153+
label = "Mask" if mask > withoutMask else "No Mask"
154+
color = (0, 255, 0) if label == "Mask" else (0, 0, 255)
155+
156+
# # Include the probability in the label
157+
# label = "{}: {:.2f}%".format(label, max(mask, withoutMask) * 100)
158+
159+
# Display the label and bounding box rectangle on the output frame
160+
cv2.putText(frame, label, (startX, startY - 10),
161+
cv2.FONT_HERSHEY_SIMPLEX, 0.45, color, 2)
162+
cv2.rectangle(frame, (startX, startY), (endX, endY), color, 2)
163+
164+
# Show the output frame
165+
cv2.imshow("Frame", frame)
166+
key = cv2.waitKey(1) & 0xFF
167+
168+
# If the `q` key was pressed, break from the loop
169+
if key == ord("q"):
170+
break
171+
172+
# Cleanup
173+
cv2.destroyAllWindows()
174+
vs.stop()
175+
except Exception as e:
176+
print(e)
177+
continue
178+
179+
window.close()
180+
181+
# # Observation:
182+
# 1)Import the necessary packages
183+
# 2)We are creating a blob and a screen for our face to be seen and highlighted
184+
# 3)Next based on the above part we make predictions if atleast 1 face is seen
185+
# 4)Read the file containing the arguments and parse the arguments like diving the data and running
186+
# 5)Load the pre-trained model for training and solving the problems
187+
# 6)Initialise the video screen and let it warm up for a time duration
188+
# 7)Keep looping until the user clicks the key q or stops running the program
189+
# 8)Keep a proper cleanup after running

‎Face_Mask_Detection 1.py‎

Lines changed: 148 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,148 @@
1+
# Python Face_Mask_Detection
2+
3+
# Import the necessary packages
4+
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
5+
from tensorflow.keras.preprocessing.image import img_to_array
6+
from tensorflow.keras.models import load_model
7+
from imutils.video import VideoStream
8+
import numpy as np
9+
import argparse
10+
import imutils
11+
import time
12+
import cv2
13+
import os
14+
15+
16+
def detect_and_predict_mask(frame, faceNet, maskNet):
17+
# Grab the dimensions of the frame and then construct a blob from it
18+
(h, w) = frame.shape[:2]
19+
blob = cv2.dnn.blobFromImage(frame, 1.0, (500, 500), (104.0, 177.0, 123.0))
20+
21+
# Pass the blob through the network and obtain the face detections
22+
faceNet.setInput(blob)
23+
detections = faceNet.forward()
24+
25+
# Initialize our list of faces, their corresponding locations and the list of predictions from our face mask network
26+
faces = []
27+
locs = []
28+
preds = []
29+
30+
# Loop over the detections
31+
for i in range(0, detections.shape[2]):
32+
# Extract the confidence (i.e., probability) associated with the detection
33+
confidence = detections[0, 0, i, 2]
34+
35+
# Filter out weak detections by ensuring the confidence is greater than the minimum confidence
36+
if confidence > args["confidence"]:
37+
# Compute the (x, y)-coordinates of the bounding box for the object
38+
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
39+
(startX, startY, endX, endY) = box.astype("int")
40+
41+
# Ensure the bounding boxes fall within the dimensions of the frame
42+
(startX, startY) = (max(0, startX), max(0, startY))
43+
(endX, endY) = (min(w - 1, endX), min(h - 1, endY))
44+
45+
# Extract the face ROI(Region of Interest), convert it from BGR to RGB channel ordering,
46+
# resize it to 224x224, and preprocess it
47+
face = frame[startY:endY, startX:endX]
48+
if face.any():
49+
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
50+
face = cv2.resize(face, (224, 224))
51+
face = img_to_array(face)
52+
face = preprocess_input(face)
53+
54+
# Add the face and bounding boxes to their respective lists
55+
faces.append(face)
56+
locs.append((startX, startY, endX, endY))
57+
58+
# Only make a predictions if at least one face was detected
59+
if len(faces) == 1:
60+
# For faster inference we'll make batch predictions on *all* faces at the same time
61+
# rather than one-by-one predictions in the above `for` loop
62+
faces = np.array(faces, dtype="float32")
63+
preds = maskNet.predict(faces, batch_size=32)
64+
65+
# Return a 2-tuple(Ordered Pair, For holding two values together of any data type) of the face locations
66+
# and their corresponding locations
67+
return locs, preds
68+
69+
70+
# Construct the argument parser(To read the file containing the arguments) and parse the arguments
71+
ap = argparse.ArgumentParser()
72+
ap.add_argument("-f", "--face", type=str, default="face_detector",
73+
help="path to face detector model directory")
74+
ap.add_argument("-m", "--model", type=str, default="mask_detector.model",
75+
help="path to trained face mask detector model")
76+
ap.add_argument("-c", "--confidence", type=float, default=0.5,
77+
help="minimum probability to filter weak detections")
78+
args = vars(ap.parse_args())
79+
80+
# Load our serialized face detector model from disk
81+
# os.path.sep.join is a method used to join components of a file with the appropriate seperator for the current os
82+
print("[INFO] loading face detector model...")
83+
prototxtPath = os.path.sep.join([args["face"], "deploy.prototxt"]) # .prototxt defines the architecture and parameters
84+
weightsPath = os.path.sep.join([args["face"],
85+
"res10_300x300_ssd_iter_140000.caffemodel"]) # .caffemodel contains the learned
86+
# weights of the model
87+
faceNet = cv2.dnn.readNet(prototxtPath, weightsPath) # cv2.dnn.readNet is a function from OpenCV's dnn module used
88+
# to read learning network models
89+
90+
# Load the face mask detector model from disk
91+
print("[INFO] loading face mask detector model...")
92+
maskNet = load_model("model.h5")
93+
94+
# Initialize the video stream and allow the camera sensor to warm up
95+
print("[INFO] starting video stream...")
96+
vs = VideoStream(src=0).start()
97+
time.sleep(2.0)
98+
99+
# Loop over the frames from the video stream
100+
while True:
101+
# Grab the frame from the threaded video stream and resize it to have a maximum width of 500 pixels
102+
# (Chose 400 pixels as a balance between reducing computational load while still maintaining enough detail
103+
# for the application's needs)
104+
frame = vs.read()
105+
frame = imutils.resize(frame, width=500)
106+
107+
# Detect faces in the frame and determine if they are wearing a face mask or not
108+
(locs, preds) = detect_and_predict_mask(frame, faceNet, maskNet)
109+
110+
# Loop over the detected face locations and their corresponding locations
111+
for (box, pred) in zip(locs, preds):
112+
# Unpack the bounding box and predictions
113+
(startX, startY, endX, endY) = box
114+
(mask, withoutMask) = pred
115+
116+
# Determine the class label and color we will use to draw the bounding box and text
117+
label = "Mask" if mask > withoutMask else "No Mask"
118+
color = (0, 255, 0) if label == "Mask" else (0, 0, 255)
119+
120+
# # Include the probability in the label
121+
# label = "{}: {:.2f}%".format(label, max(mask, withoutMask) * 100)
122+
123+
# Display the label and bounding box rectangle on the output frame
124+
cv2.putText(frame, label, (startX, startY - 10),
125+
cv2.FONT_HERSHEY_SIMPLEX, 0.45, color, 2)
126+
cv2.rectangle(frame, (startX, startY), (endX, endY), color, 2)
127+
128+
# Show the output frame
129+
cv2.imshow("Frame", frame)
130+
key = cv2.waitKey(1) & 0xFF
131+
132+
# If the `q` key was pressed, break from the loop
133+
if key == ord("q"):
134+
break
135+
136+
# Cleanup
137+
cv2.destroyAllWindows()
138+
vs.stop()
139+
140+
# # Observation:
141+
# 1)Import the necessary packages
142+
# 2)We are creating a blob and a screen for our face to be seen and highlighted
143+
# 3)Next based on the above part we make predictions if atleast 1 face is seen
144+
# 4)Read the file containing the arguments and parse the arguments like diving the data and running
145+
# 5)Load the pre-trained model for training and solving the problems
146+
# 6)Initialise the video screen and let it warm up for a time duration
147+
# 7)Keep looping until the user clicks the key q or stops running the program
148+
# 8)Keep a proper cleanup after running

0 commit comments

Comments
 (0)