NOVEL EDUCATIONAL CAMPUS PROTECTION SYSTEM

A website, an app and hardware system which will monitor if people are following covid norms and will display the statistics of number of people violating the norms in each floor of a building.

Comment

Inspiration

The entire world has come to a standstill because of the Covid-19 pandemic. The spread of the disease is increasing day by day and has shook the stability of the entire world. The whole pharma industry is striving hard to produce a vaccine on one side and the medical experts are trying their best to bring down the death rate on the other. We need to prevent it from spreading as well as get back students to colleges safely.

What it does

We can ensure people follow social distancing norms and wear masks and also view statistics of each floor to see how many are following social distancing and wearing masks in an educational campus or any MNCs. The statistics of violations can be viewed in an app and the website. Depending upon the violations in any particular floor we can take immediate strict actions. We can use the existing cameras and speakers of any particular institution to integrate it with our software. The entrance of the institution(our hardware component) will have barriers which won’t allow any person without masks to enter the institution.

How we built it

We built by creating a website which interfaces people and CCTV cameras. The website displays clear statistics of the violation of his surroundings. The mobile app also provides clear violation statistics and also have a weekly report of the violations. We also built a hardware system which ensures face mask wearing before entering into a place.

1.The live CCTV footage will be collected and sent to the college/IT campuses control room. 2.The footage will be tested by our violation software installed in the system. 3.The parts of our software is as follows:

A)Social Distance Detector: 1.The footage will be tested for human detection using the ssd_mobilenet_v1_coco model. If humans are not detected testing will be terminated in this stage. It will run this step until humans are detected.

2.If humans are detected. The model will draw a bounding box over humans. Then it will calculate the distance between the camera and humans by pixel per meter using Pythagoras theorem. This will give the distance between humans.

3.The pairs of people between which the distance is lesser than the safe threshold distance will be calculated.

4.If even one of the distances calculated is below the threshold, the trained model will  classify the place as Unsafe. If all distances are indeed above the threshold, the place is classified as Safe. 

B)Face mask detector : 1.The footage will be tested for face detection using the res10_300*300_ssd model. If the face is not detected testing will be terminated in this stage. It will run this step until a face gets detected.

2.The detected faces will be cropped.

3.A customly trained Convolutional Neural Network will be used for the detection of face masks in the frame.

4.The number of faces without facemask will be calculated. If the majority of faces are without face masks, the area of the CCTV location will be classified as Unsafe.

  1. Now, our system will check for both outputs. If the social distance detector and the face mask detector both give an output as Unsafe, then the frame is classified as Unsafe and this data along with the Floor Number and Timestamp is sent to Firebase.

  2. The Android application “COVIT” keeps continuously reading data from the Firebase cloud and updates the percentage of violations per day for a whole week in the form of a bar graph. It also shows the list of violations occurred in the descending order of time so that Users will know which place was classified as Unsafe most recently and thus can avoid that place.

7.The Android Application and the Website show a line graph of the percentage of violations in every floor. This can be further used for statistical analysis. 

8.The Hardware system is to be kept in front of entrances to check whether a person entering wears a face mask or not. The same face mask detection model is deployed on a raspberry pi using TensorflowLite. It is directly connected to a camera which is in front of the door. If the person is detected as not wearing a face mask, the system won’t allow the person inside the room/hall until the person wears a face mask.

Challenges we ran into

The main challenges are calculating social distance accurately, as different CCTV cameras have different settings calibrating them for our need is a challenge. Detecting both face mask and social distancing is a power hungry process , we need more processing power to handle many CCTVs.

Accomplishments that we're proud of

We were able to calculate the violation percentage accurately and display the same in both our website and app. Our hardware was able to detect people with no face mask and deny them access to the premises. Our application will really help people in avoiding the virus and help people to get back to normal lives safely as accurate statistics are displayed on the app and the website such that the students can avoid going to places in the college were covid norms aren't being followed as well as crowd can be scattered and strict actions can be taken where the norms aren't followed by the college management.

What we learned

We learned that our project is lot easier to implement and can be made possible in every campus and buildings easily. We also learnt to apply ML algorithms and deep learning while building this project.

What's next for NOVEL EDUCATIONAL CAMPUS PROTECTION SYSTEM

Our next step is to bring heat maps to the website , so that people can view the people density around their surroundings through the app or the website and choose to go to a particular place or floor based upon how much safety is there like if people are following covid norms or not. It will be easy to monitor all people of an institution through our website.

Built With

Share this project:

Updates