Traffic Sign Recognition Using Faster R-CNN Deep Learning Algorithm

This work proposes a framework that utilizes the deep learning algorithm of the Faster Regional Convolutional Neural Network alongside the Inception v2 extractor for the classification of traffic signs. Indian traffic speed signs of 30 and 50 have been recognized in this work. Training of neural networks is done using the TensorFlow library and learning process is conducted on “German Traffic Sign Dataset”, thus the traffic signs in the scene can be identified and classified. To study the performance of the Indian traffic sign recognition system, various experiments have been done on the two sets of Indian traffic signs, and an accuracy of up to 99% has been attained for both traffic signs after the training on the dataset used. The dataset consisted of images taken under different weather conditions, different illumination, different angles, etc. and achieving high accuracy and with such a database makes this work more significant.