zakenobi / cvhall

Projet challenge EPF Montpellier pour la détection de masque et fièvre sur système embarquer

Projet Challenge EPF : FMDE (Face Mask Detection EPF)

• Project description and Context
• How to install
• User manual
• A word on data collection

---

Project description and Context

In the context of the COVID pandemy of 2020, public places have become sensitive spots where it is important to apply health measures, including wearing a mask. This project aims to provide a Python application able to detect if people wear their mask or not using AI and measure their body temperature using a thermal camera. The application uses OpenCV and the Yolov4 algorithm for the detection and PyQt5 for the interface, it is designed to run on a Jetson Nano. Here are the different functionalities supported by the application :

• Face mask detection and display of a green square around the user if the mask is worn, a red one if it isn't
• Display of body temperature
• Collecting data (number and date of the different activations) and display of statistic charts
• Display of the project description and user manual
• Sending emails in case of detection (functional but not implemented)

---

How to install

To install the application please run the install_dependencies_jetson_nano.sh file, with these commands :

foo@bar:~$ sudo chmod +x install_dependencies_jetson_nano.sh`

foo@bar:~$ sudo ./install_dependencies_jetson_nano.sh`

The following libraries will be installed :

• PyQt5
• OpenCV
• Pandas
• matplotlib
• numpy
• adafruit
• pathlib
• mailer
• busio

---

User manual

To run the application, open the terminal and enter

python3 face_mask_detection.py

The interface is fairly easy to use, each button has a functionality :

• The green camera button runs the camera and starts the mask detection
• The red button shuts down the camera
• The button titled 'Afficher température' displays a blue square on the screen where the maximum temperature is detected
• The button 'Statistiques' displays a bar plot generated in function of the data collected by the application, the proportion between the two types of detection (the mask is worn or not) is also displayed.
• The button 'Description du projet' displays the project description and a QR Code linked to a graphic user manual

The interface is made to fit a 1280 x 720 screen and is displayed in full screen by default. To exit the app, just press alt+f4.

---

A word on data collection

To display the statistic charts, the app is recording the different detections in a text file. This file is available at app\resources\data.csv and is pre-filled with data collected during the development phase.

The syntax of the CSV file is the following (positive detections means mask worn correcly, negative detection means mask not worn) :

Date ; Time ; number of positive detection ; number of negative detections ; Sum of positive detections ; Sum of negative detections

14/06/2000 ; 08 ; 13 ; 82 ; 1594 ; 1603

14/06/2000 ; 09 ; 9 ; 8 ;;

A functionality was designed to allow the application to take screenshots in case of negative detection but was deleted from the code for confidentiality reasons.