SoilingNet: Soiling Detection on Automotive Surround-View Cameras

Cameras are an essential part of sensor suite in autonomous driving. Surround-view cameras are directly exposed to external environment and are vulnerable to get soiled. Cameras have a much higher degradation in performance due to soiling compared to other sensors. Thus it is critical to accurately detect soiling on the cameras, particularly for higher levels of autonomous driving. We created a new dataset having multiple types of soiling namely opaque and transparent. It will be released publicly as part of our WoodScape dataset [15] to encourage further research. We demonstrate high accuracy using a Convolutional Neural Network (CNN) based architecture. We also show that it can be combined with the existing object detection task in a multi-task learning framework. Finally, we make use of Generative Adversarial Networks (GANs) to generate more images for data augmentation and show that it works successfully similar to the style transfer.

I. INTRODUCTION

Autonomous driving systems are becoming mature by using a variety of different sensors. Cameras continue to be one of the key sensors as the road infrastructure is designed for human visual sensors. The first generation systems primarily used a single camera and more recently more cameras are used to get full coverage around the vehicle to handle more complex driving scenarios [5]. Horgan et al. [6] provides an overview of various visual perception tasks prior to deep learning era and Sistu et al. [12] provide an overview from a deep learning perspective.