Face Mask Detection

Face mask detection is an object detection task that detects whether people are wearing masks or not in videos. This repo includes a demo for building a face mask detector using YOLOv5 model.

Dataset

The model was trained on Face-Mask dataset which contains 853 images belonging to 3 classes, as well as their bounding boxes in the PASCAL VOC format. The classes are defined as follows:

• With mask
• Without mask
• Mask worn incorrectly

Setup

This project supports Windows, Linux, and macOS. Choose the appropriate setup method for your operating system:

Quick Setup (Automated)

For Windows (Command Prompt):

setup.bat

For Windows (PowerShell):

.\setup.ps1

Note: You may need to adjust execution policy first: Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

For Linux/macOS:

chmod +x setup.sh
./setup.sh

Manual Setup

For Windows:

# Create and activate virtual environment
python -m venv .venv
.venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Install yolov5
git clone https://github.com/ultralytics/yolov5
cd yolov5
pip install -r requirements.txt
cd ..

For Linux/macOS:

# Create and activate virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Install yolov5
git clone https://github.com/ultralytics/yolov5
cd yolov5
pip install -r requirements.txt
cd ..

For more detailed setup instructions, see SETUP.md.

Windows Users: See WINDOWS_SETUP.md for a comprehensive Windows-specific guide.

The project now includes:

• Real-time face mask detection using your computer's webcam
• Pre-trained YOLOv5 model for face mask detection
• Support for three classes: with mask, without mask, and mask worn incorrectly
• Data balancing tools to improve accuracy for underrepresented classes
• Enhanced error handling and robust configuration options

Data Preparation and Balancing

• Download Face-Mask dataset from Kaggle and copy it into datasets folder.
• Execute the following command to automatically unzip and convert the data into the YOLO format and split it into train and valid sets. The split ratio was set to 80/20%.

python prepare.py

• To address class imbalance (especially for the "mask worn incorrectly" class), you can use the data balancing script:

pip install albumentations
python balance_data.py --target-count 200

This will augment underrepresented classes to achieve more balanced training data.

Training

• Start training:

cd yolov5
python train.py --img 640 --batch 16 --epochs 100 --data ../mask_config.yaml --weights yolov5s.pt --workers 0

Inference

• If you train your own model, use the following command for inference:

python detect.py --source ../datasets/input.mp4 --weights runs/train/exp/weights/best.pt --conf 0.2

• Or you can use the pretrained model models/mask_yolov5.pt for inference as follows:

python detect.py --source ../datasets/input.mp4 --weights ../models/mask_yolov5.pt --conf 0.2

Real-time Webcam Detection

• You can use your computer's webcam for real-time face mask detection:

python webcam.py

• Additional options are available:

python webcam.py --help  # Show available options
python webcam.py --model models/mask_yolov5.pt  # Specify custom model
python webcam.py --conf-thres 0.6  # Set custom confidence threshold
python webcam.py --camera 1  # Use a different camera (if you have multiple)

• This will open your webcam and start detecting face masks in real-time. Press 'q' to quit the application.
• The application shows:
  • Green bounding box: Person wearing a mask correctly
  • Red bounding box: Person not wearing a mask
  • Yellow bounding box: Person wearing a mask incorrectly
  • FPS (Frames Per Second) in the top-left corner

Results

The following charts were obtained after training YOLOv5s with input size 640x640 on the Face Mask dataset for 100 epochs.

Precision-Recall (PR) measures the accuracy of bounding box overlap between predicted bboxes and ground truth bboxes.

Precision Curve	Recall Curve	PR Curve

The following metrics were measured on the valid dataset containing 171 images. High precision and recall scores were obtained on with mask and without mask classes. However, the performance on mask worn incorrectly class was poor due to the imbalanced data.

Class	#Labels	Precision	Recall	mAPval 0.5	mAPval 0.5:0.95
with mask	630	0.94	0.9	0.95	0.64
without mask	104	0.86	0.82	0.9	0.6
mask worn incorrectly	20	0.72	0.3	0.43	0.24
total	754	0.84	0.67	0.76	0.49

Note: You can improve the performance of the "mask worn incorrectly" class by using the data balancing script.

Ground Truths vs Predictions

Ground Truth	Prediction

Troubleshooting

If you encounter any issues, please check the following:

1. Model not found error:

   • Ensure the model file exists in the models directory
   • If training your own model, check that the training completed successfully

2. Camera access issues:

   • Ensure your webcam is properly connected
   • Try using a different camera ID with --camera flag
   • Check if another application is using the camera

3. CUDA/GPU issues:

   • Verify your CUDA installation with nvidia-smi
   • Ensure you have compatible PyTorch version for your CUDA version

For more troubleshooting tips, see SETUP.md.

Running Tests

The project uses Python's built-in unittest framework. To run all tests, execute:

python -m unittest

This will discover and run the tests located in the tests directory.

Reference

• Darknet
• YOLOv5