Abstract: A system and method for improving a video characteristic of a video stream is described. According to various implementations of the invention, a changed region between a later-in-time image frame and an earlier-in-time image frame and an unchanged region between such two image frames are determined. A new improvement to the video characteristic is determined and applied to the changed region of the later-in-time image frame. A prior improvement to the video characteristic that was determined for the earlier-in-time image frame is applied to the unchanged region of the later-in-time image frame.

Abstract: A host map system for updating a region of a host map based on sensor data received from a plurality of connected vehicles travelling in the region comprises monitoring a host map region for discrepancies, defining a geo-fence around an area of a discrepancy and setting a schedule for activing the geo-fence. When a connected vehicle enters the geo-fence, the host mapping system transmits the geo-fence to the connected vehicle and requests that it share its sensor data while travelling in the geo-fenced area. The host mapping system receives the sensor data and processes the data to update details of the host map. The host mapping system may transmit the updated map to connected vehicles travelling in the map region.

Abstract: System, methods, and other embodiments described herein relate to simulating sensor data. In one embodiment, a method includes, in response to receiving a request to generate simulated information corresponding to the sensor data, acquiring the sensor data that includes at least range information about a perceived environment. The simulated information includes one or more attributes of the sensor data that are absent from the sensor data in a current format. The method includes computing simulated information of the sensor data using a machine learning model that accepts the sensor data and labels as an input and produces the simulated information as an output. The labels identify at least objects in the perceived environment that are depicted by the sensor data. The method includes providing the simulated information with the sensor data.

Abstract: A system and method for improving a video characteristic of a video stream is described. According to various implementations of the invention, a changed region between a later-in-time image frame and an earlier-in-time image frame and an unchanged region between such two image frames are determined. A new improvement to the video characteristic is determined and applied to the changed region of the later-in-time image frame. A prior improvement to the video characteristic that was determined for the earlier-in-time image frame is applied to the unchanged region of the later-in-time image frame.

Abstract: Systems, methods, and other embodiments described herein relate to improving the classification of objects depicted in a scene. In one embodiment, a method includes generating, using an ontological detector, a type classification of a detected object according to a detector ontology of known classes. The detected object is represented as segmented data from sensor data about a surrounding environment. The method includes, in response to determining that the type classification specifies an unknown class that is not defined in the detector ontology, annotating the segmented data as unknown. The method includes providing the segmented data to specify that the type classification for the detected object is unknown.

Abstract: A vehicle behavior monitor installed in a host vehicle that monitors the behavior of other vehicles. When the behavior of the other vehicles appears to deviate from an expected trajectory of an in-vehicle map, a geo-fenced region is monitored by external sensors. The external sensor data stream is input to a pre-trained anomaly detector. The clusters from the feature space of the encoder are compared to a database of known behaviors. A confidence level is determined based on the number of vehicles which exhibit the behavior. If the confidence level is equal to or greater than a confidence level threshold, a persistence value is calculate based on the type of behavior. The behavior and the persistence value are used to update the in-vehicle map. Based on the persistence value, the update is transmitted to a map server when the host vehicle enters an area of high data connectivity or is dropped.

Abstract: A system for optimizing the performance of at least one downstream task includes one or more processors and a memory having one or more modules. The one or more modules cause the one or more processors to generate visual semantic segmentation data of a scene by a visual semantic segmentation model, generate labeled point cloud data of the scene by a vision model based on raw point cloud data of the scene and the visual semantic segmentation data, generate one or more clusters of the scene by a cluster generator model based on the labeled point cloud data, determine a clustering loss error between the one or more clusters generated by the cluster generator and one or more ground truth clusters and adjust, based on the clustering loss error, one or more model weights the visual semantic segmentation model and/or the vision model.

Abstract: Systems and methods for controlling camera settings of a camera to improve detection of faces in an uncontrolled environment are described. A first image is received from the camera, where the first image is captured by the camera at a first set of camera settings. A face is detected in the first image. The camera is adjusted to a second set of camera settings based on the detected face, where the second set of camera settings different from the first set of camera settings. A second image is received from the camera, where the second image is captured by the camera at the second set of camera settings. The face is detected in the second image. A quality metric of the face in the second image is determined where the quality metric is indicative of an image quality of the face in the second image.

Abstract: Methods and a system are described for prioritizing the upload of connected vehicle map data in a central mapping system. A central mapping unit includes a central server which receives and processes data uploads by priority order, from highest priority to lowest priority. The uploaded data is compared to central map data to determine whether the central map requires updating. When the central mapping unit determines that there is not enough information to update the central map in a region, a priority list is updated to give high priority to data collection in the region. The updated priority list is pushed to connected vehicles travelling in the region, which have external sensors for data collection in the surrounding environment. When a connected vehicle enters the region, its external sensors collect data regarding the region, which is uploaded to the central mapping unit with highest priority.

Abstract: A system and method for improving a video characteristic of a video stream is described. According to various implementations of the invention, a changed region between a later-in-time image frame and an earlier-in-time image frame and an unchanged region between such two image frames are determined. A new improvement to the video characteristic is determined and applied to the changed region of the later-in-time image frame. A prior improvement to the video characteristic that was determined for the earlier-in-time image frame is applied to the unchanged region of the later-in-time image frame.

Abstract: A host map system for updating a region of a host map based on sensor data received from a plurality of connected vehicles travelling in the region comprises monitoring a host map region for discrepancies, defining a geo-fence around an area of a discrepancy and setting a schedule for activing the geo-fence. When a connected vehicle enters the geo-fence, the host mapping system transmits the geo-fence to the connected vehicle and requests that it share its sensor data while travelling in the geo-fenced area. The host mapping system receives the sensor data and processes the data to update details of the host map. The host mapping system may transmit the updated map to connected vehicles travelling in the map region.

Abstract: A host map system for updating a region of a host map based on sensor data received from a plurality of connected vehicles travelling in the region comprises monitoring a host map region for discrepancies, defining a geo-fence around an area of a discrepancy and setting a schedule for activing the geo-fence. When a connected vehicle enters the geo-fence, the host mapping system transmits the geo-fence to the connected vehicle and requests that it share its sensor data while travelling in the geo-fenced area. The host mapping system receives the sensor data and processes the data to update details of the host map. The host mapping system may transmit the updated map to connected vehicles travelling in the map region.

Abstract: System, methods, and other embodiments described herein relate to improving clustering of points within a point cloud. In one embodiment, a method includes grouping the points into cells of a grid. The grid divides an observed region of a surrounding environment associated with the point cloud into the cells. The method includes computing feature vectors for the cells that use cell features to characterize the points in the cells and relationships between the cells. The method includes analyzing the feature vectors according to a clustering model to identify clusters for the cells. The clustering model evaluates the cells to identify which of the cells belong to common entities. The method includes providing the clusters as assignments of the points to the entities depicted in the point cloud.

Abstract: A system for optimizing the performance of at least one downstream task includes one or more processors and a memory having one or more modules. The one or more modules cause the one or more processors to generate visual semantic segmentation data of a scene by a visual semantic segmentation model, generate labeled point cloud data of the scene by a vision model based on raw point cloud data of the scene and the visual semantic segmentation data, generate one or more clusters of the scene by a cluster generator model based on the labeled point cloud data, determine a clustering loss error between the one or more clusters generated by the cluster generator and one or more ground truth clusters and adjust, based on the clustering loss error, one or more model weights the visual semantic segmentation model and/or the vision model.

Abstract: Methods and a system are described for prioritizing the upload of connected vehicle map data in a central mapping system. A central mapping unit includes a central server which receives and processes data uploads by priority order, from highest priority to lowest priority. The uploaded data is compared to central map data to determine whether the central map requires updating. When the central mapping unit determines that there is not enough information to update the central map in a region, a priority list is updated to give high priority to data collection in the region. The updated priority list is pushed to connected vehicles travelling in the region, which have external sensors for data collection in the surrounding environment. When a connected vehicle enters the region, its external sensors collect data regarding the region, which is uploaded to the central mapping unit with highest priority.

Abstract: A host map system for updating a region of a host map based on sensor data received from a plurality of connected vehicles travelling in the region comprises monitoring a host map region for discrepancies, defining a geo-fence around an area of a discrepancy and setting a schedule for activing the geo-fence. When a connected vehicle enters the geo-fence, the host mapping system transmits the geo-fence to the connected vehicle and requests that it share its sensor data while travelling in the geo-fenced area. The host mapping system receives the sensor data and processes the data to update details of the host map. The host mapping system may transmit the updated map to connected vehicles travelling in the map region.

Abstract: A vehicle behavior monitor installed in a host vehicle that monitors the behavior of other vehicles. When the behavior of the other vehicles appears to deviate from an expected trajectory of an in-vehicle map, a geo-fenced region is monitored by external sensors. The external sensor data stream is input to a pre-trained anomaly detector. The clusters from the feature space of the encoder are compared to a database of known behaviors. A confidence level is determined based on the number of vehicles which exhibit the behavior. If the confidence level is equal to or greater than a confidence level threshold, a persistence value is calculate based on the type of behavior. The behavior and the persistence value are used to update the in-vehicle map. Based on the persistence value, the update is transmitted to a map server when the host vehicle enters an area of high data connectivity or is dropped.

Abstract: System, methods, and other embodiments described herein relate to improving the classification of objects depicted in a scene. In one embodiment, a method includes generating, using an ontological detector, a type classification of a detected object according to a detector ontology of known classes. The detected object is represented as segmented data from sensor data about a surrounding environment. The method includes, in response to determining that the type classification specifies an unknown class that is not defined in the detector ontology, annotating the segmented data as unknown. The method includes providing the segmented data to specify that the type classification for the detected object is unknown.

Abstract: Systems and methods for controlling camera settings of a camera to improve detection of faces in an uncontrolled environment are described. A first image is received from the camera, where the first image is captured by the camera at a first set of camera settings. A face is detected in the first image. The camera is adjusted to a second set of camera settings based on the detected face, where the second set of camera settings different from the first set of camera settings. A second image is received from the camera, where the second image is captured by the camera at the second set of camera settings. The face is detected in the second image. A quality metric of the face in the second image is determined where the quality metric is indicative of an image quality of the face in the second image.

Abstract: System, methods, and other embodiments described herein relate to improving clustering of points within a point cloud. In one embodiment, a method includes grouping the points into cells of a grid. The grid divides an observed region of a surrounding environment associated with the point cloud into the cells. The method includes computing feature vectors for the cells that use cell features to characterize the points in the cells and relationships between the cells. The method includes analyzing the feature vectors according to a clustering model to identify clusters for the cells. The clustering model evaluates the cells to identify which of the cells belong to common entities. The method includes providing the clusters as assignments of the points to the entities depicted in the point cloud.