Robust Lane Detection via Filter Estimator and Data Augmentation

Abstract

Lane detection is an important topic in the self-driving system. Having a stable lane detection system will assist the self-driving cars to make decisions in order to bring a more comfortable and safe driving environment to the driver. In this paper, we use a network architecture composed of Encoder-Decoder with a Feature Shift Aggregator between them to make the prediction more comprehensive; through our dataset, we found that some problems such as glitches occur when changing lanes. In this regard, we use Data Augmentation and Filter respectively to solve the problem. Finally, the network achieves the result accuracy rate of SOTA on the TuSimple dataset.