Chapter 30: Face Mask Detection

“When vision meets responsibility—models that see and protect.”

---

😷 Why Face Mask Detection?

The COVID-19 pandemic highlighted the need for real-time face mask detection in public safety systems. Detecting whether a person is wearing a mask can:

• Enhance compliance in public places
• Assist automated access control
• Provide insights for health monitoring systems

This task is a real-world application of binary object classification layered on top of face detection.

---

Problem Breakdown

Detect faces in an image

Classify each face:

• with_mask
• without_mask

This is different from simply classifying the whole image—it’s region-specific classification.

---

Dataset

We’ll use the Face Mask Detection Dataset, which contains:

• Labeled images of people with and without masks
• Optionally: bounding boxes (if using detection-first approach)

Alternatively, for simplicity, use a folder-structured dataset:

dataset/
  ├── with_mask/
  └── without_mask/

---

Implementation Steps

Step 1: Load and Preprocess Images

from tensorflow.keras.preprocessing.image import ImageDataGenerator

datagen = ImageDataGenerator(rescale=1./255, validation_split=0.2)

train_gen = datagen.flow_from_directory(
    'dataset',
    target_size=(128, 128),
    batch_size=32,
    class_mode='binary',
    subset='training'
)

val_gen = datagen.flow_from_directory(
    'dataset',
    target_size=(128, 128),
    batch_size=32,
    class_mode='binary',
    subset='validation'
)

---

Step 2: Build a CNN Model

from tensorflow.keras import layers, models

model = models.Sequential([
    layers.Conv2D(32, (3,3), activation='relu', input_shape=(128,128,3)),
    layers.MaxPooling2D(2,2),

    layers.Conv2D(64, (3,3), activation='relu'),
    layers.MaxPooling2D(2,2),

    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(1, activation='sigmoid')  # Binary classification
])

---

Step 3: Compile and Train

model.compile(optimizer='adam',
              loss='binary_crossentropy',
              metrics=['accuracy'])

history = model.fit(train_gen, validation_data=val_gen, epochs=10)

---

Step 4: Make Predictions

import numpy as np
from tensorflow.keras.preprocessing import image

img = image.load_img('test.jpg', target_size=(128, 128))
img_array = image.img_to_array(img) / 255.0
img_array = np.expand_dims(img_array, axis=0)

prediction = model.predict(img_array)[0][0]
print("With mask" if prediction < 0.5 else "Without mask")

---

Optional: Use OpenCV for Real-Time Detection

Integrate with your webcam:

import cv2

# Load model and Haar Cascade for face detection
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')

cap = cv2.VideoCapture(0)

while True:
    _, frame = cap.read()
    faces = face_cascade.detectMultiScale(frame, 1.3, 5)

    for (x,y,w,h) in faces:
        face = frame[y:y+h, x:x+w]
        face_resized = cv2.resize(face, (128,128)) / 255.0
        prediction = model.predict(np.expand_dims(face_resized, axis=0))[0][0]
        label = "With Mask" if prediction < 0.5 else "No Mask"
        color = (0,255,0) if prediction < 0.5 else (0,0,255)
        cv2.rectangle(frame, (x,y), (x+w, y+h), color, 2)
        cv2.putText(frame, label, (x,y-10), cv2.FONT_HERSHEY_SIMPLEX, 0.8, color, 2)

    cv2.imshow('Face Mask Detector', frame)
    if cv2.waitKey(1) == 27: break  # Esc to exit

cap.release()
cv2.destroyAllWindows()

---

Deployment Tip

Use:

• TensorFlow Lite for mobile deployment
• FastAPI + OpenCV for web camera APIs
• Hugging Face Spaces for demo apps

---

Summary

In this chapter, you built a binary image classifier to detect face masks using TensorFlow and CNNs. You also explored how to:

• Train a model with folder-labeled images
• Use real-time webcam feeds for predictions
• Visualize and deploy your results

---