Covid Defender

According to data obtained by the World Health Organization, The pandemic of COVID-19has severely impacted the world and has now infected over fifty million people worldwide. Wearing face masks and following safe social distancing are two of the improved safety protocols that must be followed public

Project Title

Covid Defender

Project Area of Specialization

Artificial Intelligence

Project Summary

According to data obtained by the World Health Organization, The pandemic of COVID-19has severely impacted the world and has now infected over fifty million people worldwide. Wearing face masks and following safe social distancing are two of the improved safety protocols that must be followed publically places to prevent the spread of the virus. To form a safe environment that contributes to public safety, we propose an efficient computer vision-based approach focused on the real-time automated monitoring of individuals to detect both safe social distancing and face masks publicly places by implementing the model on Raspberry Pi4 to observe activity and detect violations through the camera. After detection of the breach, the Raspberry pi4 sends an alert signal to regulate the Centre at the state station house and also gives the alarm to the public. This proposed system based on a modern deep learning algorithm and mixed with geometric techniques for building a strong modal that covers three aspects of detection, tracking, and validation. Thus, the proposed system favors society by saving time and helps in preventing the spread of coronavirus. It can be implemented effectively in the current situation to examine persons publicly gatherings, shopping malls, colleges, etc. Automated inspection reduces manpower to examine the general public and would be a useful tool to reduce the spread of this communicable disease for many countries in the world

Project Objectives

As Covid 19 came the whole world shut down for months and still, they are in the Lockdown process. By implementing this project in Schools, Universities, Offices and etc, we can assure our GOVERNMENT OFFICIALS that how we can follow strictly SOPs by using that technology and can save thousands of lives and prevent corona cases and can live our lives normally and regularize life again.The solution has the potential to significantly reduce violations by real-time interventions,  so the proposed system would improve public safety by saving time and help to prevent the spread of coronavirus. This solution can be used in places like schools, shopping malls, airports, etc

Project Implementation Method

We developed 3-stage model systems that perform social distancing detection, tracking, mask detection, and temperature detection through camera s. The system can be integrated and applied to all types of CCTV surveillance cameras with any resolution from normal to Full-HD, with real-time performance. This system uses a deep learning algorithm and a computer vision to automatically monitor people in public places with a camera integrated with a raspberry pi4 and to detect people with masks or no masks.

In this process we use a raspberry pi camera which feeds video from real-time are streamed using Raspberry pi and then these frames are converted to grayscale to improve speed and accuracy because in grayscale that picture has fewer pixels which contain fewer data and trim other data which system not required and then send to the model for further processing inside raspberry pi4. We have used the MobileNetV2 architecture and YOLO model for detection as MobileNetV2 provides a huge cost advantage compared to the normal 2D CNN mode 

To calculate the distance among two persons first the distance of person from camera is calculated using triangle correspondence technique, we calculate perceived focal length of camera, we suppose person distance D from camera and person's actual height H=190cms and with SSD person detection pixel height  P of the person is identified using the bounding box coordinates. To measure the center point, C(x, y), of the bounding box for the detected person, midpoint equation is used C(x, y)=(xmin+xmax2,ymin+ymax2) To measure the distance, C1(xmin , ymin) and C2(xmax , ymax), between each of the detected person in the frame, distance equation is used d(C1?C2)=(xmax?xmin)2+(ymax?ymin)2.

Using these values, the focal length of the camera can be measured using the formula below. F11 = (P1 x D1) / H1

Then we use the real person's height H, the person's pixel height P, and the camera's focal length F to measure the person's distance from the camera. The distance from the camera can be obtained using the following.

D12 = (H1 x 1F) / P1

The developed system is using one of the high performings trained SSD model for face detection with mobile Net V2 architecture as the backbone to create a lightweight model that is accurate and computationally efficient, making it easier to implement the model to a raspberry pi. We used different face datasets of about 3500 images in mask and no mask. These images are used to train a deep learning model that identifies the input image into the mask and no mask categories using the pattern. Deploying our model to edge devices for automatic monitoring of public places could reduce the burden of physical monitoring, which is why we choose to use this architecture. This system can be integrated with edge device for use in airports, railway stations, offices, schools, and public places to ensure that public safety guidelines are followed

Benefits of the Project

We developed a system that uses computer vision and YOLO model architecture to help maintain a secure environment and provide individuals protection by monitoring public places to avoid the spread of the COVID-19 virus and try to reduce manpower by minimizing their physical surveillance work in public areas where surveillance can be done through a camera with raspberry pi4 in real-time.

This system will operate systematically in the current situation when the lockdown will ease after the second wave and it helps to track public places easily in an automated manner. We have addressed in depth the tracking of social distancing and the identification of face masks that help to ensure human health. The implementation of this solution was successfully tested in real-time by deploying the model in raspberry pi4. The solution has the potential to significantly reduce violations by real-time interventions,  so the proposed system would improve public safety by saving time and help to prevent the spread of coronavirus. This solution can be used in places like schools, shopping malls, airports, etc

Technical Details of Final Deliverable

Social Distancing Monitoring - We experimented with two different deep CNN-based object detectors CNN and YOLOv4 Figure 1 shows the distancing monitoring results using Faster R-CNN. The detector performances are given in Table 1. As can be seen in the table, both detectors achieved real-time performance. The precision of the distance measurement between pedestrians is also affected by the pedestrian detection algorithm. The YOLO algorithm is also able to detect the half body of the pedestrian. If the social distance is not preserved, the system generates a warning and sends an alert to authorities with a face image.

Mask Detection -The developed system uses a custom data set consisting of thousands of face images with different types of face masks which are used for the training of our model. The model first detects all persons in the range of cameras and shows a green bounding box around each person who is far from each other after that model conducts a test on the identification of social distances maintained in a public place if persons breaching social distance norms bounding box color changes to red for those persons and simultaneously face mask detection is achieved by showing bounding boxes on the identified person's face with mask or non-mask labeled and also showing scores. The system detects the social distancing and masks with a precision score of 95.1% with a precision value of 0.87 and FPS = 20.07.

Temperature Detection - For temperature detection, we use thermal cameras to detecting people with fever-like symptoms in high-traffic areas such as hospital entrances, shopping centers, and offices, and potentially mass-attendance sporting events and we have achieved above 90 percent accuracy. Results suggest thermal cameras are a perfect solution for preventing the spread of the coronavirus.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Health

Other Industries

Others

Core Technology

Others

Other Technologies

Artificial Intelligence(AI)

Sustainable Development Goals

Good Health and Well-Being for People

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com