Detection of Road Accident Using Dense Convolutional Network

Accidents are the leading cause of deaths and have not changed much, especially in developing countries, with growing improvements in safety measures. This study uses the data mining techniques to recognize the most pivotal factors that determine the accident severity. The delay in medical attention may also increase the number of fatalities of the victims involved in accidents. To overcome these challenges, a system is introduced that can identify accidents on roads and inform the next control room. The camera module of the system shall be placed in accident-prone areas. In the event of an accident, the system shall identify the occurrence and send the message to the next control room instantly. This system operates based on deep learning methodologies, especially convolutional neural networks. This system implementation is likely to reduce the number of casualties significantly. This system makes use of the Dense Convolutional Network called DenseNet, which has the potential to process a video frame by frame, with the help of this deep learning model trained the frames of accidents and non-accidents are identified effectively. Convolutional neural networks have shown itself to be a fast and precise method for classifying images. DCNN-based image classifiers have reached accuracies more than 95% on relatively small datasets and require much less preprocessing compared to other image classification algorithms.