India has around one and a half lakh persons die due to road accidents per year. One of the main reasons for this number is no timely availability of help. Automatic accident detection can shorten the response time of rescue agencies and vehicles around accidents to improve rescue efficiency and traffic safety level. The ability to detect and track vehicle can be used in applications like monitoring road accidents. The proposed system uses YOLOv5 deep learning algorithm to detect the vehicles from the real time CCTV surveillance video. The primary focus of the system is to build a model that detects various class of vehicle by using custom dataset. The dataset consists of 1000 images with various condition such as rainfall, low visibility, luminosity, and weather conditions. The proposed framework uses YOLOv5 to detect vehicle with improved efficiency in real time object detection system. This model can further be extended to analyse and classify accidents using 3D Convolutional Neural Network based on the severity of the accident and to alert the nearest hospital.