An Acoustic based Roadside Symbols Detection and Identification using Faster RCNN and SSD

Currently, the safety and security of people on the road has been an important concern area. Every day, in newspapers and television a lot of news could be seen of mishaps on roads because of negligence. This research has been carried out for providing a safe environment for drivers, visually impaired people. This paper illustrates the experiment conducted on Roadside traffic symbols to increase the efficiency and accuracy. The two algorithms named Regional proposal based Algorithm that is Faster RCNN and Regression Based Algorithm that is Single Short Multibox Detector are used respectively. After the detection and identification of the traffic symbols a sound is produced which speaks out the recognized symbol name to the user. The comparison between these algorithms is made to find which algorithm's performance is better based upon different parameters The different graphs for loss function, learning rate, accuracy, training and testing time are a few parameters for both the algorithms which shows that the Single Shot Multibox is better than Faster RCNN.