This document proposes a system to detect social distancing violations using computer vision and deep learning algorithms. The system would identify individuals in video frames using a YOLOv3 model, calculate distances between detected individuals, and classify the risk level based on social distancing guidelines. It transforms frames into a bird's eye view to standardize distance measurements. The proposed system aims to help monitor social distancing and slow the spread of COVID-19 by identifying groups that are too close together. It achieved 92.8% precision in social distancing classification during testing.