Covid Voilation detection

Project Title

Covid Voilation detection

Project Area of Specialization Artificial IntelligenceProject Summary

The ongoing COVID-19 corona virus outbreak has caused a global disaster with its deadly spreading. Due to the absence of effective remedial agents and the shortage of immunizations against the virus, population vulnerability increases. In the current situation, as there are no vaccines available; therefore, social distancing is thought to be an adequate precaution (norm) against the spread of the pandemic virus. The risks of virus spread can be minimized by avoiding physical contact among people. The purpose of this work is, therefore, to provide a deep learning platform for social distance tracking using an overhead perspective. The framework uses the YOLOv3 object recognition paradigm to identify humans in video sequences. The transfer learning methodology is also implemented to increase the accuracy of the model. In this way, the detection algorithm uses a pre-trained algorithm that is connected to an extra trained layer using an overhead human data set. The detection model identifies peoples using detected bounding box information. Using the Euclidean distance, the detected bounding box centroid's pairwise distances of people are determined. To estimate social distance violations between people, we used an approximation of physical distance to pixel and set a threshold. A violation threshold is established to evaluate whether or not the distance value breaches the minimum social distance threshold. In addition, a tracking algorithm is used to detect individuals in video sequences such that the person who violates/crosses the social distance threshold is also being tracked. Experiments are carried out on different video sequences to test the efficiency of the model. Findings indicate that the developed framework successfully distinguishes individuals who walk too near and breaches/violates social distances; also, the transfer learning approach boosts the overall efficiency of the model. The accuracy of 92% and 98% achieved by the detection model without and with transfer learning, respectively. The tracking accuracy of the model is 95%.

Project Objectives

In this work, a deep learning-based social distance monitoring framework is presented using an overhead perspective. The pre-trained YOLOv3 paradigm is used for human detection. As a person's appearance, visibility, scale, size, shape, and pose vary significantly from an overhead view, the transfer learning method is adopted to improve the pre-trained model's performance. The model is trained on an overhead data set, and the newly trained layer is appended with the existing model. To the best of our knowledge, this work is the first attempt that utilized transfer learning for a deep learning-based detection paradigm, used for overhead perspective social distance monitoring. The detection model gives bounding box information, containing centroid coordinates information. Using the Euclidean distance, the pairwise centroid distances between detected bounding boxes are measured. To check social distance violations between people, an approximation of physical distance to the pixel is used, and a threshold is defined. A violation threshold is used to check if the distance value violates the minimum social distance set or not. Furthermore, a centroid tracking algorithm is used for tracking peoples in the scene. Experimental results indicated that the framework efficiently identifies people walking too close and violates social distancing; also, the transfer learning methodology increases the detection model's overall efficiency and accuracy. For a pre-trained model without transfer learning, the model achieves detection accuracy of 92% and 95% with transfer learning. The tracking accuracy of the model is 95%. The work may be improved in the future for different indoor and outdoor environments. Different detection and tracking algorithms might be used to help track the person or people who are violating or breaches the social distancing threshold.

Project Implementation Method

• Overhead perspective for human detection and tracking. The overhead perspective offers a better field of view and overcomes the issues of occlusion, thereby playing a key role in social distance monitoring to compute the distance between peoples. It might help overcome computation, communication load, energy consumption, human resource, and installation costs
   
• This work aims to present a deep learning-based social distance monitoring framework for the public campus environment from an overhead perspective. A deep learning model, i.e., YOLOv3 (You Only Look Once)  is applied for human detection. The current model (pre-trained on frontal or normal view data sets) is initially tested on the overhead data set. Transfer learning is also used to improve the efficiency of the detection model.
• The data set is composed of frontal and side view images of limited people. The work is also extended for the monitoring of facial masks. The drone camera and the YOLOv3 algorithm help identify the social distance and monitor people from the side or frontal view in public wearing masks.
• We used Euclidean distance and calculated the distance between each detected bounding box of peoples. Following computing centroid distance, a predefined threshold is used to check either the distance among any two bounding box centroids is less than the configured number of pixels or not. If two people are close to each other and the distance value violates the minimum social distance threshold. The bounding box information is stored in a violation set, and the color of the bounding box is updated/changed to red.
• YOLOv3 is used for human detection as it improves predictive accuracy, particularly for small-scale objects. The main advantage is that it has adjusted network structure for multi-scale object detection. Furthermore, for object classification, it uses various independent logistic rather than softmax.

• During human identification, the input frame is divided into a region of S×S, also called grid cells. These cells are related to bounding box estimation and class probabilities. It predicts the probability of whether the center of the person bounding box is in the grid cell or not:

  Conf(p)=Pr(p)×IOU(pred,actual)

• The localization loss estimates the failures in the predicted bounding box sizes and locations. The bounding box containing the detected object, i.e., a person, is added. 

•  Centroid tracking algorithm is used to track the detected people in the video sequence. The tracking algorithm also helps to keep track of people who are violating the social distance threshold. At the output, the model displays information about the total number of social distancing violations.

Benefits of the Project

The transfer learning methodology is applied to improve the accuracy of the detection model. Using an overhead data set, the model is additionally trained using 500 sample frames. The epoch size 40 and batch size 64 is set for training of the model.

The main advantage is that it has adjusted network structure for multi-scale object detection. Furthermore, for object classification, it uses various independent logistic rather than softmax.

• To present a deep learning-based social distance monitoring framework using an overhead view perspective.
• • To deploy pre-trained YOLOv3 for human detection and computing their bounding box centroid information. In addition, a transfer learning method is applied to enhance the performance of the model. The additional training is performed with overhead data set, and the newly trained layer is appended to the pre-trained model.
• • In order to track the social distance between individuals, the Euclidean distance is used to approximate the distance between each pair of the centroid of the bounding box detected. In addition, a social distance violation threshold is specified using a pixel to distance estimation.
• • Utilizing a centroid tracking algorithm to keep track of the person who violates the social distance threshold.
• • To assess the performance of pre-trained YOLOv3 by evaluating it on an overhead data set. The output of the detection framework is assessed with and without the transfer learning. Furthermore, the model performance is also compared with other deep learning models.

Technical Details of Final Deliverable

1. We will use Good IP camera for live detection of mask and violation
2. we have use sperate camera for heat fever detection.
3. we have use jetson nano board for processing and use 64 gb M-card 
4. All the detection will be shown on web application and notifiation on webapplication as well as mobile.
5. we will set camers on stand so detection will give proper result.
6. used Usb hub for wiring 
7. High qaulity cameras used

Final Deliverable of the Project HW/SW integrated systemCore Industry HealthOther Industries Education , IT , Medical , Agriculture Core Technology Artificial Intelligence(AI)Other Technologies Internet of Things (IoT)Sustainable Development Goals Decent Work and Economic Growth, Reduced Inequality, Sustainable Cities and Communities, Responsible Consumption and Production, Peace and Justice Strong InstitutionsRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60300
Ip Camera 3mp 360	Equipment	2	6000	12000
heat detection sensor lmx x35a	Equipment	1	4000	4000
Nodemcu	Equipment	1	1300	1300
Jetson nano board 2gb latest	Equipment	1	29000	29000
memory card 64gb	Equipment	1	1000	1000
heat detection camera 2mp	Equipment	1	6000	6000
wiring and Usb hub	Equipment	1	2000	2000
camera stand	Miscellaneous	1	5000	5000