Abstract: In various examples, one or more DNNs may be used to detect landmarks (e.g., lane lines) and regress a representation of their shape. A DNN may be used to jointly generate classifications of measured 3D points using one output head (e.g., a classification head) and regress a representation of one or more fitted shapes (e.g., polylines, circles) using a second output head (e.g., a regression head). In some embodiments, multiple DNNs (e.g., a chain of multiple DNNs or multiple stages of a DNN) are used to sequentially generate classifications of measured 3D points and a regressed representation of the shape of one or more detected landmarks. As such, classified landmarks and corresponding fitted shapes may be decoded and used for localization, navigation, and/or other uses.

Abstract: In various examples, perception of landmark shapes may be used for localization in autonomous systems and applications. In some embodiments, a deep neural network (DNN) is used to generate (e.g., per-point) classifications of measured 3D points (e.g., classified LiDAR points), and a representation of the shape of one or more detected landmarks is regressed from the classifications. For each of one or more classes, the classification data may be thresholded to generate a binary mask and/or dilated to generate a densified representation, and the resulting (e.g., dilated, binary) mask may be clustered into connected components that are iteratively: fitted a shape (e.g., a polynomial or Bezier spline for lane lines, a circle for top-down representations of poles or traffic lights), weighted, and merged. As such, the resulting connected components and their fitted shapes may be used to represent detected landmarks and used for localization, navigation, and/or other uses.