Real-time Indian TSR using MDEffNet

Abstract

Traffic sign recognition (TSR) is an important aspect of Intelligent transport systems (ITS). The lack of architectures that are flexible across multiple datasets are rarely there. As such this work attempts to use an efficient Deep neural network (DNN) architecture and implement it on an Indian traffic sign dataset IRSDBv1.0. IRSDBv1.0 is the first publicly available Indian datset, to the best our knowledge. The performance of MDEffNet is studied and analyzed.