Main Article Content
Accidents have been a significant reason for deaths in India. Over 80% of accident related incident happen not because of the mishap itself yet the absence of ideal assistance arriving at the mishap casualties. In interstates where the traffic is truly light and speedy a mishap casualty could be left unattended for quite a while. The purpose is to make a framework which would distinguish a mishap dependent on the live feed of video from a CCTV camera introduced on an interstate. The thought is to take each casing of a video and run it through a profound learning - Recurrent neural organization model which has been prepared to arrange edges of a video into mishap or non-mishap in progressive way. Progressive Recurrent Neural Networks has demonstrated to be a quick and precise way to deal with arranges pictures. H-RNN based picture classifiers have given precisions of over 95% for relatively more modest datasets and require less preprocessing when contrasted with other picture characterizing calculations. Contrasted with conventional RNNs, H-RNN is more reasonable to video extraction, since it can misuse long fleeting reliance among outlines, then; the calculation activities are fundamentally decreased.