Social distancing is the only way to control the spread of COVID-19,so peoples aremaintain the distance in public places. Recently, AI teams are created Social Distancing Tools to contol the covid by using the concepts of Computer Vision. This project proposes a methodology to detect social distance using deep learning for the evaluation of distance between people to mitigate the impact of this corona virus pandemic. The detection tool was developed to alert people to keep safe distance among each other by evaluating a video input feed. The video frame from the file .avi was given as input, and the object detection pre-trained model based on the YOLOv3 algorithm was employed for pedestrian detection. Then, the video frame was transformed into top-down view for distance measurement from the 2D plane. In existing system, image acquisition is carried out by first selecting the video file and split them into frames. Then the images are taken for pedestrian detection. For better results, images can be resized but not resized in existing system. If the distance less than the acceptable distance between any two individuals, will be indicated with red lines that serve as precautionary warnings. The YOLO trained on the COCO dataset which consists of 80 labels including human or pedestrian class. In this work, the only box coordinates, object confidence and pedestrian object class from detection result in the YOLO model were used for pedestrian detection. In existing system have some drawbacks, Multiple group of persons closest with them is not considered. Half body human image cannot be labeled as person. In proposed system, image acquisition is carried out by first selecting the video file and split them into frames. Then the images are taken for pedestrian detection. For better results, images can be resized here. If the distance less than the acceptable distance between any two individuals, will be indicated with red lines that serve as precautionary warnings. The YOLO trained on the COCO dataset which consists of 80 labels like „person?, etc including human or pedestrian class. In this work, the only box coordinates, object confidence and pedestrian object class from detection result in the YOLO model were used for pedestrian detection. Confidence value for label “person” is adjusted here with default value set as 0.5.In proposed system overcome the drawbacks of existing system, Multiple group of persons closest with them is not considered. Half body human image cannot be labeled as person and overlapped images can be considered.