Face Mask Detector

A real-time face mask detection system using a convolutional neural network (CNN). This project detects whether a person is wearing a mask or not from live video feed or images.

---

🚀 Features

• Real-Time Detection: Detects faces and classifies them as "With Mask" or "Without Mask".
• Data Augmentation: Enhances training dataset with transformations like rotation, zoom, and flipping.
• User-Friendly Interface: Visualizes detection results with bounding boxes and color-coded labels.
• Customizable: Easy to modify for different datasets or detection tasks.

---

📂 Project Structure

Face-Mask-Detector/
├── train.py          # Script to train the mask detection model
├── test.py           # Script to test the model with live video feed
├── model-010.h5      # Pre-trained model file
├── train/            # Training dataset directory
├── test/             # Testing dataset directory
├── README.md         # Project documentation

---

🛠️ Prerequisites

Ensure you have the following installed:

1. Python 3.6+
2. Keras
3. TensorFlow
4. OpenCV
5. Numpy

Install dependencies with:

pip install -r requirements.txt

---

⚙️ Setup

1. Clone the Repository:

   git clone https://github.com/yourusername/Face-Mask-Detector.git
   cd Face-Mask-Detector

2. Prepare Dataset:

   • Place your training dataset in the train/ folder.
   • Place your testing dataset in the test/ folder.
   • The dataset should have two subdirectories: mask/ and no_mask/.

3. Train the Model: Run the train.py script to train the model:

   python train.py

4. Test the Model: Use the test.py script to run real-time detection:

   python test.py

---

📊 Model Details

The CNN architecture includes:

• Two Convolutional Layers with 100 filters each.
• MaxPooling layers to reduce spatial dimensions.
• Dropout for regularization.
• Fully connected Dense layers for classification.

Loss function: binary_crossentropy
Optimizer: adam
Metrics: accuracy

---

🧪 Data Augmentation

The training data is augmented with:

• Rotation: up to ±40 degrees.
• Zoom: range of 20%.
• Shifts: Horizontal and vertical by 20%.
• Flipping: Horizontal flips.

This improves model robustness and prevents overfitting.

---

📈 Training

The model trains for 10 epochs with ImageDataGenerator for augmentation.

Example command:

python train.py

---

🖥️ Usage Instructions

1. Run the detection system:

   python test.py

2. Press Esc to exit the live detection window.

---

📂 Dataset Example

Your dataset structure should look like:

train/
├── mask/
│   ├── image1.jpg
│   ├── image2.jpg
│   ...
├── no_mask/
│   ├── image1.jpg
│   ├── image2.jpg
│   ...

test/
├── mask/
├── no_mask/

---

🤝 Contributing

Contributions are welcome! Feel free to submit a PR or open an issue.

---

🛡️ License

This project is licensed under the MIT License.

---

📧 Contact

For any queries or issues, reach out to:

• Email: vikas.sharma005@protonmail.com
• GitHub: vikasharma005