A deep neural network(s) (DNN) may be used to detect objects from sensor data of a three dimensional (3D) environment. For example, a multi-view perception DNN may include multiple constituent DNNs or stages chained together that sequentially process different views of the 3D environment. An example DNN may include a first stage that performs class segmentation in a first view (e.g., perspective view) and a second stage that performs class segmentation and/or regresses instance geometry in a second view (e.g., top-down). The DNN outputs may be processed to generate 2D and/or 3D bounding boxes and class labels for detected objects in the 3D environment. As such, the techniques described herein may be used to detect and classify animate objects and/or parts of an environment, and these detections and classifications may be provided to an autonomous vehicle drive stack to enable safe planning and control of the autonomous vehicle.

An apparatus, method and computer program product are provided for determining a state of visibility for a road object. In one example, the apparatus receives temporal data, calculates an orientation of a light source with respect to a road object using the temporal data, and predicts a state of visibility for the road object based on the orientation of the light source. In another example, the apparatus determines an artificial light source associated with a road object, receives attribute data associated with the artificial light source, determines a state of the artificial light source using the attribute data, and predicts a state of visibility for the road object based on the state of the artificial light source.

An information processing device includes processing circuitry. The processing circuitry obtain target information that indicates at least one of a distance to a target object or a position of the target object. The processing circuitry generate, based on the target information, map information of a space including a plurality of areas, the map information indicating presence or absence of the target object in a first area included in the plurality of areas, and a detailed position of the target object in the first area.

The present disclosure relates to a control device, and a control method, a program, and a mobile body that enable efficient search for surrounding information when it is in an own position indefinite state. When it is in an own position indefinite state, on the basis of an own position, obstacle position information around oneself, and information of a surface sensing possible range of a surface sensing unit including a stereo camera for determining the own position, information of a surface-sensed area of an obstacle is recorded, and a search route is planned on the basis of the information of the surface-sensed area of the obstacle. The present technology can be applied to a multi-legged robot, a flying body, and an in-vehicle system that autonomously move according to a mounted computer.

In various examples, sensor data may be adjusted to represent a virtual field of view different from an actual field of view of the sensor, and the sensor data—with or without virtual adjustment—may be applied to a stereographic projection algorithm to generate a projected image. The projected image may then be applied to a machine learning model—such as a deep neural network (DNN)—to detect and/or classify features or objects represented therein.

To determine a path through a pose configuration space, trajectories of poses may be evaluated in parallel based at least on translating the trajectories along at least one axis of the pose configuration space (e.g., an orientation axis). A trajectory may include at least a portion of a turn having a fixed turn radius. Turns or turn portions that have the same turn radius and initial orientation can be translatively shifted along and processed in parallel along the orientation axis as they are translated copies of each other, but with different starting points. Trajectories may be evaluated based at least on processing variables used to evaluate reachability as bit vectors with threads effectively performing large vector operations in synchronization. A parallel reduction pattern may be used to account for dependencies that may exist between sections of a trajectory for evaluating reachability, allowing for the sections to be processed in parallel.

A method for recognizing an object includes reading in a first image signal that represents a first camera image recorded by a mono camera, and a second image signal that represents a second camera image recorded by the mono camera. First pixels situated on an image line of the first camera image are selected from the first image signal. Second pixels are identified from the second image signal, the second pixels corresponding to the first pixels. A flux signal is formed using the first pixels and the second pixels, the flux signal representing an optical flux profile for the first pixels situated along the image line. The flux profile represented by the flux signal is segmented into a plurality of segments, each of which represents a plane in the vehicle surroundings. An object signal that represents a recognized object is determined, using the plurality of segments.

A method and system of operating a vehicle includes a first sensor generating first sensor data for an object comprising a first bounding box from a first sensor. The first sensor data comprising a first confidence score. A second sensor generates second sensor data for the object comprising a second bounding box from a second sensor different than the second sensor. The second sensor data comprises a second confidence score. A bounding box circuit is programmed to generate a third confidence score for the object based on the first sensor data and the second sensor data and utilize the first sensor data, the second sensor data and the third confidence score to control operation of a vehicle system.

An obstacle positioning method, device and terminal are provided. The method includes determining installation positions of at least two detectors on a vehicle, and respective detection areas of the detectors, determining an overlapping area of the detection areas of the detectors, and if determining that an obstacle is located in the overlapping area, determining a position of the obstacle according to the installation positions of the detectors forming the overlapping area. By changing the number and positions of detectors installed on an unmanned vehicle, a plurality of overlapping areas of the detection areas of the detectors are obtained, the distribution of obstacles around the vehicle are optimally identified, so that the unmanned vehicle makes reasonable driving plans based on an accurate surrounding obstacle environment.

A deep neural network(s) (DNN) may be used to detect objects from sensor data of a three dimensional (3D) environment. For example, a multi-view perception DNN may include multiple constituent DNNs or stages chained together that sequentially process different views of the 3D environment. An example DNN may include a first stage that performs class segmentation in a first view (e.g., perspective view) and a second stage that performs class segmentation and/or regresses instance geometry in a second view (e.g., top-down). The DNN outputs may be processed to generate 2D and/or 3D bounding boxes and class labels for detected objects in the 3D environment. As such, the techniques described herein may be used to detect and classify animate objects and/or parts of an environment, and these detections and classifications may be provided to an autonomous vehicle drive stack to enable safe planning and control of the autonomous vehicle.