Safe Road AI: Real-Time Smart Accident Detection for Multi-Angle Crash Videos using Deep Learning Techniques and Computer Vision
Senthil G. A, R. V Lakshmi Priya, S Geerthik, G Karthick, R Lavanya
Abstract
On national highways, in isolated places, or at night, many accidents go unreported. If the accident victim receives medical attention soon after the incident, about 40% of fatalities from accidents can be avoided. Among the causes of traffic deaths are increased traffic volumes, excessive speeding, careless and intoxicated driving, driver weariness, inadequate road infrastructure, and the presence of animals on roadways in forest locations. Road accident fatalities as a percentage of all deaths globally have grown by 2.2%, according to the World Health Organization (WHO). Every year, traffic accidents claim the lives of almost 1.35 million individuals. Emergency medical help is often delayed in traffic incidents that lead to the deaths of people. The novelty of research proposes IoT sensor network, GSM, GPS, that we’ll employ Deep Learning (DL) algorithms like CNN (Convolutional Neural Network) to better effectively identify auto accidents. The sensor camera captures images through IoT sensor networks and computer vision. The image resizing method is employed in data preparation to prepare the datasets. Here, deep learning (DL) algorithms will be used to train these datasets, and a model file will be produced. The police will be notified as soon as an accident is discovered via the notification. The prediction technique can accurately forecast an automobile collision when given an input image. Consequently, this study uses a deep learning system to effectively determine vehicle accidents with an accuracy of about 95%.