Road images augmentation with synthetic traffic signs using neural networks
This work addresses the issue of rare traffic sign recognition in computer vision, which is an incremental improvement for autonomous driving systems.
The paper tackled the problem of rare traffic sign detection and classification by using synthetic training data generated through GAN-based methods to embed signs realistically into real photos, resulting in improved accuracy for both classifier and detector when combined with real data.
Traffic sign recognition is a well-researched problem in computer vision. However, the state of the art methods works only for frequent sign classes, which are well represented in training datasets. We consider the task of rare traffic sign detection and classification. We aim to solve that problem by using synthetic training data. Such training data is obtained by embedding synthetic images of signs in the real photos. We propose three methods for making synthetic signs consistent with a scene in appearance. These methods are based on modern generative adversarial network (GAN) architectures. Our proposed methods allow realistic embedding of rare traffic sign classes that are absent in the training set. We adapt a variational autoencoder for sampling plausible locations of new traffic signs in images. We demonstrate that using a mixture of our synthetic data with real data improves the accuracy of both classifier and detector.