Disclosed is a system to add photorealistic detail and motion to an image based on a first material property associated with a first set of data points of an incomplete first object, and a second material property associated with a second set of data points of an incomplete second object in the image. The system may generate first artificial data points amongst the first set of data points that completes a first arrangement defined for the first material property, and may generate second artificial data points amongst the second set of data points that completes a second arrangement defined for the second material property. The system may then output an enhanced image of the completed first object based on first set of data points and the first artificial data points, and of the completed second object based on the second set of data points and the second artificial data points.

Disclosed is a system and associated methods for generating and rendering a polyhedral point cloud that represents a scene with multi-faceted primitives. Each multi-faceted primitive stores multiple sets of values that represent different non-positional characteristics that are associated with a particular point in the scene from different angles. For instance, the system generates a multi-faceted primitive for a particular point of the scene that is captured in first capture from a first position and a second capture from a different second position. Generating the multi-faceted primitive includes defining a first facet with a first surface normal oriented towards the first position and first non-positional values based on descriptive characteristics of the particular point in the first capture, and defining a second facet with a second surface normal orientated towards the second position and second non-positional values based on different descriptive characteristics of the particular point in the second capture.

The subject matter of this specification can be implemented in, among other things, methods, systems, computer-readable storage medium. A method can include receiving, by a processing device, image data including one or more image frames indicative of a current state of a meal preparation area. The processing device determines a first quantity of a first ingredient disposed within a first container based on the image data. The processing device determines a meal preparation procedure associated with the first ingredient based on the first quantity. The processing device causes a notification indicative of the meal preparation procedure to be displayed on a graphical user interface (GUI).

A method for generating a parallax-tolerant panoramic image includes obtaining a point cloud captured by a depth sensor, the point cloud representing a support structure bearing a set of objects; obtaining a set of images of the support structure and the set of objects, the set of images captured by an image sensor from a plurality of positions alongside a length of the support structure; generating a mesh structure using the point cloud, the mesh structure including a plurality of patches and representing a surface of the support structure and the set of objects; for each patch in the mesh structure, selecting an image from the set of images and projecting the selected image to the mesh patch; and generating an orthographic projection of the mesh structure onto a shelf plane of the support structure.

Described herein are systems, methods, and non-transitory computer readable media for performing an alignment between a first vehicle sensor and a second vehicle sensor. Two-dimensional (2D) data indicative of a scene within an environment being traversed by a vehicle is captured by the first vehicle sensor such as a camera or a collection of multiple cameras within a sensor assembly. A three-dimensional (3D) representation of the scene is constructed using the 2D data. 3D point cloud data also indicative of the scene is captured by the second vehicle sensor, which may be a LiDAR. A 3D point cloud representation of the scene is constructed based on the 3D point cloud data. A rigid transformation is determined between the 3D representation of the scene and the 3D point cloud representation of the scene and the alignment between the sensors is performed based at least in part on the determined rigid transformation.

Embodiments of the present disclosure relate to methods and apparatuses for outputting information and calibrating a camera. The method may include: acquiring a first image, a second image, and a third image, the first image being an image photographed by a to-be-calibrated camera, the second image being a high-precision map image including a target area indicated by the first image, and the third image being a reflectance image including the target area; fusing the second image and the third image to obtain a fused image; determining a matching point pair based on points selected by a user in the first image and the fused image; and calibrating the to-be-calibrated camera based on coordinates of the matching point pair.

The subject matter of this specification can be implemented in, among other things, methods, systems, computer-readable storage medium. A method can include receiving, by a processing device, image data having one or more image frames indicative of a state of a meal preparation area. The method may further include, determining, based on the image data, a first feature characterization of a first meal preparation item associated with the state of the meal preparation area. The method may further include determining that the first feature characterization does not meet object classification criteria for a set of object classifications. The method may further include causing a notification indicating the first meal preparation item and one of an object classification or a classification status corresponding to the first meal preparation item on a graphical user interface (GUI).

In one embodiment, a method for extracting point cloud features for use in localizing an autonomous driving vehicle (ADV) includes selecting a first set of keypoints from an online point cloud, the online point cloud generated by a LiDAR device on the ADV for a predicted pose of the ADV; and extracting a first set of feature descriptors from the first set of keypoints using a feature learning neural network running on the ADV, The method further includes locating a second set of keypoints on a pre-built point cloud map, each keypoint of the second set of keypoints corresponding to a keypoint of the first set of keypoint; extracting a second set of feature descriptors from the pre-built point cloud map; and estimating a position and orientation of the ADV based on the first set of feature descriptors, the second set of feature descriptors, and a predicted pose of the ADV.

An occlusion detection system for an autonomous vehicle is described herein, where a signal conversion system receives a three-dimensional sensor signal from a sensor system and projects the three-dimensional sensor signal into a two-dimensional range image having a plurality of pixel values that include distance information to objects captured in the range image. A localization system detects a first object in the range image, such as a traffic light, having first distance information and a second object in the range image, such as a foreground object, having second distance information. An occlusion polygon is defined around the second object and the range image is provided to an object perception system that excludes information within the occlusion polygon to determine a configuration of the first object. A directive is output by the object perception system to control the autonomous vehicle based upon occlusion detection.

“Feature points” in “point clouds” that are visible to multiple respective cameras (i.e., aspects of objects imaged by the cameras) are reported via wired and/or wireless communication paths to a compositing processor which can determine whether a particular feature point “moved” a certain amount relative to another image. In this way, the compositing processor can determine, e.g., using triangulation and recognition of common features, how much movement occurred and where any particular camera was positioned when a latter image from that camera is captured. Thus, “overlap” of feature points in multiple images is used so that the system can close the loop to generate a SLAM map. The compositing processor, which may be implemented by a server or other device, generates the SLAM map by merging feature point data from multiple imaging devices.