DA4AD: End-to-End Deep Attention-based Visual Localization for Autonomous Driving

Abstract

We present a visual localization framework aided by novel deep attention aware features for autonomous driving that achieves centimeter level localization accuracy. Conventional approaches to the visual localization problem rely on handcrafted features or human-made objects on the road. They are known to be either prone to unstable matching caused by severe appearance or lighting changes, or too scarce to deliver constant and robust localization results in challenging scenarios. In this work, we seek to exploit the deep attention mechanism to search for salient, distinctive and stable features that are good for longterm matching in the scene through a novel end-to-end deep neural network. Furthermore, our learned feature descriptors are demonstrated to be competent to establish robust matches and therefore successfully estimate the optimal camera poses with high precision. We comprehensively validate the effectiveness of our method using a freshly collected dataset with high-quality ground truth trajectories and hardware synchronization between sensors. Results demonstrate that our method achieves a competitive localization accuracy when compared to the LiDAR-based localization solutions under various challenging circumstances, leading to a potential low-cost localization solution for autonomous driving. IntroductionLocalization is a fundamental task in a self-driving car system. To exploit high definition (HD) maps as priors for robust perception and safe motion planning, this requires the localization system to reach centimeter-level accuracy [2].Despite many decades of research, building a long-term, precise and reliable localization system using low-cost sensors, such as automotive and consumergrade GPS/IMU and cameras, is still an open-ended and challenging problem. Compared to the LiDAR, that cameras are passive sensors meaning that they are more susceptible to appearance changes caused by varying lighting conditions or changes in viewpoint. It is known that handcrafted point features (DIRD [26,28],

Attention-SLAM: A Visual Monocular SLAM Learning from Human Gaze

“…Salient DSO [20] combines a saliency model and image segmentation model with Direct Sparse Odometry (DSO) [21]. Zhou et al [22] utilize saliency information to extract and match features. This method enhances the performance of DSO.…”