Systems, methods, and non-transitory computer-readable media can perform operations comprising determining visual geometry detections (e.g., lane line detections) associated with geometry corresponding with a map; aligning the visual geometry detections with the geometry based on transformations associated with selected degrees of freedom; and determining a pose of a vehicle based on alignment of the visual geometry detections with the geometry.

This application disclose a method for training a model performed at a computing device. The method includes: acquiring a template image and a test image; invoking a first object recognition model to process a feature of a tracked object in the template image to obtain a first reference response, and a second object recognition model to process the feature in the template image to obtain a second reference response; invoking the first model to process a feature of a tracked object in the test image to obtain a first test response, and the second model to process the feature to obtain a second test response; tracking the first test response to obtain a tracking response of the tracked object; and updating the first object recognition model based on differences between the first and second reference responses, that between the first and second test responses, and that between a tracking label and the tracking response.

Aspects of the disclosure provide for generating a visualization of a three-dimensional (3D) world view from the perspective of a camera of a vehicle. For example, images of a scene captured by a camera of the vehicle and 3D content for the scene may be received. A virtual camera model for the camera of the vehicle may be identified. A set of matrices may be generated using the virtual camera model. The set of matrices may be applied to the 3D content to create a 3D world view. The visualization may be generated using the 3D world view as an overlay with the image, and the visualization provides a real-world image from the perspective of the camera of the vehicle with one or more graphical overlays of the 3D content.

The present application discloses a method and apparatus for determining a route, a device and a computer storage medium, and relates to the field of big data. An implementation includes: acquiring route description information input by a user, and acquiring a route contour with the route description information; matching the route contour in road network data to obtain the route matched with the route contour, so as to generate recommended routes, wherein this operation specifically includes: extracting intersection points in the route contour, each of which is formed by intersecting at least two lines; selecting one of the intersection points, and querying the road network data using a corresponding angle sequence of the selected intersection point in the route contour, so as to obtain intersections matched with the selected intersection point as candidate intersections; traversing the candidate intersections, fixedly mapping the selected intersection point to the positions of the candidate intersections, equally scaling the route contour in the road network data, and recording mapped road information if all lines of the route contour are mapped onto connected roads; and generating the recommended routes with the recorded road information.

A method, apparatus, and system for estimating a moving speed of a detected pedestrian at an autonomous driving vehicle (ADV) is disclosed. A pedestrian is detected in a plurality of frames of point clouds generated by a LIDAR device installed at an autonomous driving vehicle (ADV). In each of at least two of the plurality of frames of point clouds, a minimum bounding box enclosing points corresponding to the pedestrian excluding points corresponding to limbs of the pedestrian is generated. A moving speed of the pedestrian is estimated based at least in part on the minimum bounding boxes across the at least two of the plurality of frames of point clouds. A trajectory for the ADV is planned based at least on the moving speed of the pedestrian. Thereafter, control signals are generated to drive the ADV based on the planned trajectory.

Implementations generally perform pose reconstruction by tracking for video analysis. In some implementations, a method includes obtaining a plurality of videos of at least one subject performing at least one action in an environment. The method further includes tracking the at least one subject across at least two cameras. The method further includes reconstructing a 3-dimensional (3D) model of the at least one subject based on the plurality of videos and the tracking of the at least one subject.

A moving object detection system detecting moving objects based on a captured image comprising: cameras capturing images to generate image data; image recognition circuitries each performing image recognition on the image data generated by each camera to detect a moving object in a captured image indicated by the image data; and an information management apparatus managing image data supplied from each camera to a corresponding image recognition circuitry. The information management apparatus includes a communication interface performing data communication with the cameras and the image recognition circuitries, and a controller managing congestion information on the number of moving objects localized within a range of a captured image indicated by image data for each camera, and controlling the communication interface to supply the image data generated by the cameras to each image recognition circuitry, distributing the image data according to the congestion information.

A crowd motion simulation method is provided based on real crowd motion videos. The method includes framing the videos and storing the framed videos into continuous high-definition images, generating a crowd density map of each image, and accurately positioning an individual in each density map to obtain the accurate position of each individual. The method also includes correlating the positions of each individual in different images to form a complete motion trajectory, and extracting motion trajectory data; and quantifying motion trajectory data, defining training data and data labels, and calculating data correlation. The method further includes building a deep convolutional neural network, and inputting the motion trajectory data for training to learn crowd motion behaviors; and randomly placing a plurality of simulation individuals in a two-dimensional space, testing a prediction effect of the deep convolutional neural network, adjusting parameters for simulation, and drawing a crowd motion trajectory.

A speed estimation system includes: a detection module configured to: detect an object on a surface in an image captured using a camera; and generate a bounding box around the object; a Jacobian module configured to generate a Jacobian for the object based on the bounding box; and a speed module configured to determine a speed that the object is traveling on the surface based on the Jacobian.

Systems and techniques described herein relate to techniques for improving image detection. In some examples, aspects relate to systems and techniques for improving image detection by performing tracking of objects within captured image frames. A process can include obtaining, from an image capture device, a first image frame including an object. The process can further include determining, using an object detector, an object validation score associated with detection of the object in the first image frame, and determining the object validation score is less than a validation threshold. Based on the object validation score being less than the validation threshold, the process can include tracking the object for one or more image frames received subsequent to the first image frame.