Automatic Detection of Social Distance Violation from Real Time Web Camera Using Open CV with Deep Learning

Abstract

In current days there is a great demand for social distancing due to covid19 pandemic. As we all know lot of survey reports tell that by maintaining proper social distancing one can able to reduce the spread of covid 19.In this current article, we presents a methodology for social distancing detection using OpenCV with deep learning  to identify the distance between people to mitigate the impact of this coronavirus pandemic. The proposed model was designed to alert people to maintain a safe distance with each other by evaluating a video feed. In order to test the model, we try to collect video sequences collected from CCTV cameras and then try to apply some pre-trained CNN models on that input video. Here we try to apply the YOLOv3 algorithm  for identifying the persons who are walking on the road.Once the pedestrian detection is completed, now the video file is converted into a top-down view for distance measurement from the 2D plane. If the distance between any pair of people is less than expected length then it is marked in red color frame and identified as social distance violation is clearly seen. If the distance between any pair of people seems more than the expected distance, than those pair of pedestrians are marked with blue or green color. In order to test the accuracy of our current model,we try to test the model on a pre-recorded video of pedestrians walking on the street. The result shows that the proposed method is able to determine the social distancing measures between multiple people in the video. This same model can be deployed as a tool in future for real time applications to detect the social distance violations.