Computer vision and machine learning application implemented in Python to detect masks in one or more faces. Supports detection from static images and video streams from webcam.
The easiest way to use this program is to run the gui.py file (currently, only supported on UNIX systems). If you are using the Windows operating systems run the code (detect_mask_image.py or detect_mask_video.py) directly from the terminal.
The dependencies to run this script are:
- Tensorflow
- Keras
- OpenCV (cv2)
- Skicit-learn (sklearn)
- Matplotlib
- Numpy
- Imutils
- Tkinter
Everyone recognizes the importance of wearing a mask to reduce the spread of SARS-CoV-2, the virus that causes COVID-19. However, instead of letting people monitor each other, the intention of this project is to develop a computer program that would be able to analyze images or real-time video stream from a webcam and detect whether people are wearing a mask or not. This mask recognition software does not hurt privacy because the program does not actually identify people. The software is trained according to two sets of images: one that teaches the algorithm how to recognize a face (“face detection”) and another that defines how to recognize a face with a mask (“mask recognition”). The machine learning algorithm does not identify the face in a way that can link it to a specific person, because it does not use any type of training that links faces to identities.
This project consists of four Python scripts:
train_mask_detector.py: Takes the input dataset and creates the detector model using the MobileNetV2 convolutional neural network architecture. A training report image containing the accuracy/loss graph is also created when this script is run.detect_mask_image.py: Search for facial masks in static images.detect_mask_video.py: Search for facial masks in every frame of the webcam video stream.gui.py: Combines all three previously cited scripts into one graphical user interface program.
You'll need to install the necessary packages so that the script can run without any problems.
Before installing the dependencies, please make sure you have python3 installed on your machine. But since almost all Linux distros come with Python pre-installed you probably won't need to perform this step. After that, on the Linux terminal, type the following commands as root:
sudo pip3 install tensorflow
sudo pip3 install keras
sudo pip3 install sklearn
sudo pip3 install opencv-contrib-python
sudo pip3 install matplotlib
sudo pip3 install numpy
sudo pip3 install imutils
sudo pip3 install tk
Since Windows doesn't come with Python pre-installed, you'll need to install Python if you haven't already. It is recommended to install the stable release.
When in the installation, be sure to check the option Add Python to environment variables.
Then, after installation, go to -> "Start" and type Manage App Execution Aliases. Go to it and turn off 'Python'.
After that, install the following dependecies by typing the following commands one by one on the Command Prompt (CMD):
python -m pip install tensorflow
python -m pip install keras
python -m pip install sklearn
python -m pip install opencv-contrib-python
python -m pip install matplotlib
python -m pip install numpy
python -m pip install imutils
python -m pip install tk
macOS comes with Python, but is a deprecated version that is no longer supported. So, you should install a newer version of Python. It is recommended to install the stable release.
Then, type the following command on the terminal to install the necessary packages:
pip install tensorflow
pip install keras
pip install sklearn
pip install opencv-contrib-python
pip install matplotlib
pip install numpy
pip install imutils
pip install tk
The gui.py file can be used to perform all tasks.
This model is already trained, so you probably won't need to perform this step. However, if you still insist or want to retrain the model, the gui.py file can be used. First, on the GUI, go to the Settings tab and click 'Browse' to select the path to the dataset you wish to use to train the model. After that, go to the Training tab and click 'Train model'. Another alternative to train the model is to use one of the following commands depending on your system's OS:
On Linux, type:
python3 train_mask_detector.py --dataset dataset
However, if you're using Windows or macOS type the following command:
python train_mask_detector.py --dataset dataset
After issuing this command, the terminal will continuously prompt current information about the training process including accuracy, loss, number of epoch and remaining time.
After the training process is done, inside the root folder, an image file will be created containing the performance metrics of the model. Also, some metrics like precision, recall f1-score and support will be displayed in the terminal.
To test the model using static images, apart from using the GUI (under the Image tab), which is recommended since it makes the process fairly simple, the following commands can also be issued in the terminal as an example. For Linux, type the following on the terminal:
python3 detect_mask_image.py --image examples/ex1
On Windows and macOS, use the following command:
python detect_mask_image.py --image examples/ex1
More images were then used to continue to test the model. Some of the results of these tests are shown below.
After assuring that the detect_mask_image.py script is working properly, it is time to test the detect_mask_video.py script to see if it also preforms without issues. Once again, to do that the gui.py file can be used (in the Video tab) or depending on your system's OS, the following commands can be issued in the terminal.
On Linux, type:
python3 detect_mask_video.py
On Windows and macOS, use the following:
python detect_mask_video.py
