Traffic Sign Classification Using Generative Adversarial Network

The development of autonomous vehicles has seen rapid progress in the last few years due to its enormous potential. Autonomous driving technologies depend on accurate discrimination of traffic signs. To improve the safety and efficiency of autonomous vehicles, this paper aims to create a reliable model that can recognize traffic signs from pictures taken by onboard cameras. For this task, the use of the generative adversarial network (GAN), more specifically, an autoencoder/generator, and a discriminator/classifier are investigated in this work. The suggested approach uses a discriminator network for classification after a convolutional autoencoder that artificially creates suitable training images from real photos. By harnessing the adversarial training process, the autoencoder can produce more diverse features, thereby enhancing overall classification accuracy. The effectiveness of our approach in accurately classifying traffic signs with high-performance metrics is reported through extensive experimentation and evaluation. The proposed GAN model demonstrates competitive performance and achieves classification accuracy of 97.06% where the accuracy of the second-best model is 96.39%.