As per the Road safety status report published by MHRD, India road traffic accidents are the 13th largest contributor to deaths in the country over the last decade. In India, highways contribute to only 2% of the roads. But, the highways alone account for 36% of fatalities. A quick first response to accidents would reduce the mortality rate significantly. Automatic detection of accidents and alerting the concerned authorities by leveraging technologies such as Computer vision, edge computing and advanced communication systems would significantly reduce the mortality rate. In this work, an automatic accident detection mechanism using deep learning and a scheme for notifying the concerned authorities is proposed. The proposed work uses the already-existing CCTV cameras in highway networks to provide a novel method of accident detection. The suggested system analyses real-time video feeds for the detection of possible accidents using deep learning algorithms and edge caching technology. The main data sources in this system are CCTV cameras that are strategically placed throughout the highways. The edge caching nodes at the intersections process the video streams from these cameras in real-time. A deep learning model is trained to identify patterns, crashes, and unusual vehicle actions, that are suggestive of accidents. This enhances analysis of real-time traffic feed and prediction of crashes automatically. When a potential crash is detected, emergency services are immediately notified allowing for prompt response and help. Use of edge caching infrastructure minimizes latency and the strain on centralized servers, it improves the scalability and responsiveness of the system. Furthermore, the suggested method is more workable and economical because the deployment and cost issues related to outfitting cars with individual sensors are addressed by the usage of CCTV cameras.