Analysis of Classifier Training on Synthetic Data for Cross-Domain Datasets

A major challenges of deep learning (DL) is the necessity to collect huge amounts of training data. Often, the lack of a sufficiently large dataset discourages the use of DL in certain applications. Typically, acquiring the required amounts of data costs considerable time, material and effort. To mitigate this problem, the use of synthetic images combined with real data is a popular approach, widely adopted in the scientific community to effectively train various detectors. In this study, we examined the potential of synthetic data-based training in the field of intelligent transportation systems. Our focus is on camera-based traffic sign recognition applications for advanced driver assistance systems and autonomous driving. The proposed augmentation pipeline of synthetic datasets includes novel augmentation processes such as structured shadows and gaussian specular highlights. A well-known DL model was trained with different datasets to compare the performance of synthetic and real image-based trained models. Additionally, a new, detailed method to objectively compare these models is proposed. Synthetic images are generated using a semi-supervised errors-guide method which is also described. Our experiments showed that a synthetic image-based approach outperforms in most cases real image-based training when applied to cross-domain test datasets (+10% precision for GTSRB dataset) and consequently, the generalization of the model is improved decreasing the cost of acquiring images.

I. Introduction

Transportation systems around the world are becoming technologically more complex every day. The number of vehicles in the transportation system is also increasing and transportation channels are getting congested. Thus, ensuring safe driving has become a major global challenge. Advanced driver assistant systems (ADAS) are a major step toward achieving safe driving because they (i) support the driver by providing assistance in fundamental skills and (ii) serve as end-to-end systems that support self-driving vehicles. These trends are reflected in the automobile market as the automotive industry is regularly introducing ADAS in new vehicles to minimize accidents and thus increase safety. ADAS include blind spot detection (BSD), lane departure warning (LDW), collision avoidance systems (CAS), traffic sign recognition (TSR), driver drowsiness detectors (DDD), etc. For example, TSR systems analyze images of a forward looking camera to detect and recognize traffic signs. This can be used either to inform or warn the driver about relevant current traffic rules or to feed the decision system of an autonomous driving system.