In the recent scenario, there is a drastic improvement in transportation, infrastructure, and communication technology which increases the number of commercial as well as non-commercial vehicles. Therefore, there is also an increase in the number of accident incidence. This ultimately results in a high death rate due to a road accident. More than half of accident incidence results in death due to delayed medical aid to the victim. If medical aid or services received at the proper time, then the victim may survive. With the application of machine learning processes and communication advancements, there is scope for the development of a more accurate system. In this chapter, a model is presented based on IoT devices that can sense and predict the pre-accident/pre-collision state and generates an alarm message about the collision is going to occur. This model is designed to extracts image/video features to determine the possibility of occurrence of a collision. This model is also efficient for post-collision.