Abstract: A device may process video data, with a feature extraction model, to generate features, and may process a customer identifier, with an embedding layer, to generate an input. The device may optimize model weights for a classifier model and a customizer model, and may process the features, with the classifier model, to generate general predictions. The device may process the features, the input, and the general predictions, with the customizer model, to generate customer predictions, and may calculate first errors for the general predictions. The device may calculate second errors for the customer predictions, and may train the classifier model and the feature extraction model with the first errors and the optimized model weights. The device may train the customizer model and the embedding layer with the second errors and the optimized model weights. The device may implement the trained classifier model, feature extraction model, customizer model, and embedding layer.

Abstract: In some implementations, a device may determine a plurality of driving conditions associated with an image of a road scene based on providing a set of features associated with the image to a plurality of processing layers of a model. Each processing layer, of the plurality of processing layers, may determine, in parallel, a respective driving condition of the plurality of driving conditions and may comprise a plurality of sequential linear layers including a first sequential, linear layer comprising a first quantity of neurons corresponding to a quantity of features included in the set of features and computing resources of the device and a last sequential, linear layer comprising a second quantity of neurons that is based on a task associated with determining the respective driving condition. The device may perform one or more actions based on the plurality of driving conditions.

Abstract: A device may receive accelerometer data and video data for a vehicle and may identify bounding boxes and object classes for objects near the vehicle. The device may identify tracks for the objects and may filter out tracks that are not associated with vehicles or vulnerable road users to generate one or more tracks or an indication of no tracks. The device may generate a collision cone identifying a drivable area of the vehicle to identify objects more likely to be involved in a collision and may filter out tracks from the one or more tracks, based on the bounding boxes, and to generate a subset of tracks or another indication of no tracks. The device may determine scores for the subset of tracks and may identify a track of the subset of tracks with a highest score. The device may perform actions based on the identified track.

Abstract: In some implementations, an adverse environment detection system may receive an image of a road scene associated with a vehicle. The adverse environment detection system may determine a set of features associated with the image based on providing the image to an initial portion of a model. The adverse environment detection system may determine a first condition associated with the image based on providing the set of features to a first processing layer of the model, a second condition associated with the image based on providing the set of features to a second processing layer of the model, and a third condition associated with the image based on providing the set of features to a third processing layer of the model. The first processing layer, the second processing layer, and the third processing layer may process the set of features in parallel.

Abstract: A device may receive schedule data identifying schedules of appointments for drivers of vehicles and may receive location data identifying geographical locations of the vehicles. The device may receive traffic data identifying traffic conditions, and may process the received data, with a first model, to determine status data identifying estimated statuses of the appointments. The device may process the received data, with a second model, to generate sets of isochrones for destinations of the appointments, and may calculate, based on the sets of isochrones and the location data, estimated travel time data identifying estimated travel times for the appointments. The device may process the status data, the estimated travel time data, and the schedule data, with a third model, to calculate estimated time of arrival data identifying estimated times of arrival for the appointments, and may perform actions based on the estimated time of arrival data.

Abstract: A routing optimization system may determine a plurality of routes based on one or more changes to an initial route. Each change generates a respective route of the plurality of routes. The routing optimization system may store route information for each route. The route information, for each route, includes cost information identifying a cost associated with the route. The route information is stored in a respective entry of a data structure. The routing optimization system may determine that a particular route, associated with a lowest cost out of costs associated with the plurality of routes, is to be selected from the plurality of routes. The routing optimization system may determine whether an entry, of a plurality of entries associated with the plurality of routes, is empty. The routing optimization system may select the particular route based on the plurality of entries after determining that the entry is not empty.

Abstract: A load planning platform may generate, according to one or more loading rules, a preliminary packing solution that simulates placing unpacked items, in a set of items, into a container. The load planning platform may generate a set of packing solutions by applying one or more available moves to the preliminary packing solution. The load planning platform may select a final packing solution from the set of packing solutions based on one or more optimization criteria associated with the container and the set of items. The load planning platform may provide access to a three-dimensional rendering of the final packing solution that differentiates each item in the set of items based on a sequence in which the set of items are to be unloaded from the container.

Abstract: A device may receive sensor data and video data associated with a vehicle, and may process the sensor data, with a rule-based detector model, to determine whether a probability of a vehicle accident satisfies a first threshold. The device may preprocess acceleration data of the sensor data to generate calibrated acceleration data, and may process the calibrated acceleration data, with an anomaly detector model, to determine whether the calibrated acceleration data includes anomalies. The device may filter the sensor data to generate filtered sensor data, and may process the filtered sensor data and anomaly data, with a decision model, to determine whether the probability of the vehicle accident satisfies a second threshold. The device may process the filtered sensor data, the anomaly data, and the video data, with a machine learning model, to determine whether the vehicle accident has occurred, and may perform one or more actions.

Abstract: A device may receive accelerometer data and video data for a vehicle and may identify bounding boxes and object classes for objects near the vehicle. The device may identify tracks for the objects and may filter out tracks that are not associated with vehicles or vulnerable road users to generate one or more tracks or an indication of no tracks. The device may generate a collision cone identifying a drivable area of the vehicle to identify objects more likely to be involved in a collision and may filter out tracks from the one or more tracks, based on the bounding boxes, and to generate a subset of tracks or another indication of no tracks. The device may determine scores for the subset of tracks and may identify a track of the subset of tracks with a highest score. The device may perform actions based on the identified track.

Abstract: In some implementations, an adverse environment detection system may receive an image of a road scene associated with a vehicle. The adverse environment detection system may determine a set of features associated with the image based on providing the image to an initial portion of a model. The adverse environment detection system may determine a first condition associated with the image based on providing the set of features to a first processing layer of the model, a second condition associated with the image based on providing the set of features to a second processing layer of the model, and a third condition associated with the image based on providing the set of features to a third processing layer of the model. The first processing layer, the second processing layer, and the third processing layer may process the set of features in parallel.

Abstract: A device may receive accelerometer data and video data for a vehicle and may identify bounding boxes and object classes for objects near the vehicle. The device may identify tracks for the objects and may filter out tracks that are not associated with vehicles or vulnerable road users to generate one or more tracks or an indication of no tracks. The device may generate a collision cone identifying a drivable area of the vehicle to identify objects more likely to be involved in a collision and may filter out tracks from the one or more tracks, based on the bounding boxes, and to generate a subset of tracks or another indication of no tracks. The device may determine scores for the subset of tracks and may identify a track of the subset of tracks with a highest score. The device may perform actions based on the identified track.

Abstract: A device may receive accelerometer data and video data for a vehicle and may identify bounding boxes and object classes for objects near the vehicle. The device may identify tracks for the objects and may filter out tracks that are not associated with vehicles or vulnerable road users to generate one or more tracks or an indication of no tracks. The device may generate a collision cone identifying a drivable area of the vehicle to identify objects more likely to be involved in a collision and may filter out tracks from the one or more tracks, based on the bounding boxes, and to generate a subset of tracks or another indication of no tracks. The device may determine scores for the subset of tracks and may identify a track of the subset of tracks with a highest score. The device may perform actions based on the identified track.

Abstract: A device may obtain video data associated with a driving event involving a first vehicle. The device may determine a vanishing point associated with the video data and may construct a cone of impact of the first vehicle based on the vanishing point. The device may detect a second vehicle within the cone of impact and may analyze the subset of video frames to determine a distance between the first vehicle and the second vehicle. The device may determine a speed of the first vehicle during a time period associated with a subset of video frames. The device may determine a headway score, representative of a severity associated with the first vehicle being within a proximity threshold of the second vehicle during the time period, based on the distance and the speed. The device may determine an occurrence of a tailgating event based on the headway score.

Abstract: A system described herein may provide a technique for the identification of solutions, such as routing solutions based on a starting point and a destination, based on Key Performance Indicators (“KPIs”) associated with models that are included in, or are associated with, candidate solutions. A solution request may specify a set of KPI preferences, based on which certain KPIs may be more heavily weighted than others in the identification of a suitable solution. Multiple solutions may be identified, with measures of comparative differences or similarities identified. A requestor may be able to make an informed decision between multiple candidate solutions based on the identification of the comparative differences.

Abstract: A device may obtain video data associated with a driving event involving a first vehicle. The device may determine a vanishing point associated with the video data and may construct a cone of impact of the first vehicle based on the vanishing point. The device may detect a second vehicle within the cone of impact and may analyze the subset of video frames to determine a distance between the first vehicle and the second vehicle. The device may determine a speed of the first vehicle during a time period associated with a subset of video frames. The device may determine a headway score, representative of a severity associated with the first vehicle being within a proximity threshold of the second vehicle during the time period, based on the distance and the speed. The device may determine an occurrence of a tailgating event based on the headway score.

Abstract: A device may receive sensor data and video data associated with a vehicle, and may process the sensor data, with a rule-based detector model, to determine whether a probability of a vehicle accident satisfies a first threshold. The device may preprocess acceleration data of the sensor data to generate calibrated acceleration data, and may process the calibrated acceleration data, with an anomaly detector model, to determine whether the calibrated acceleration data includes anomalies. The device may filter the sensor data to generate filtered sensor data, and may process the filtered sensor data and anomaly data, with a decision model, to determine whether the probability of the vehicle accident satisfies a second threshold. The device may process the filtered sensor data, the anomaly data, and the video data, with a machine learning model, to determine whether the vehicle accident has occurred, and may perform one or more actions.

Abstract: A device can receive, from a client device, a request to modify a schedule that includes information identifying the set of deliveries that the fleet of vehicles is to perform. The device can generate a set of modified schedules using information included in the request and one or more routing techniques. The device can determine, for each modified schedule, of the set of modified schedules, one or more scores for ranking the set of modified schedules, such as a projected delivery time score, an operational cost score, and/or an overall score. The device can provide the set of modified schedules that have been scored to the client device. The device can receive a modified schedule that has been selected by the client device. The device can deploy the selected modified schedule to instruct the fleet of vehicles to perform an updated set of deliveries associated with the selected modified schedule.

Abstract: A device may receive schedule data identifying schedules of appointments for drivers of vehicles and may receive location data identifying geographical locations of the vehicles. The device may receive traffic data identifying traffic conditions, and may process the received data, with a first model, to determine status data identifying estimated statuses of the appointments. The device may process the received data, with a second model, to generate sets of isochrones for destinations of the appointments, and may calculate, based on the sets of isochrones and the location data, estimated travel time data identifying estimated travel times for the appointments. The device may process the status data, the estimated travel time data, and the schedule data, with a third model, to calculate estimated time of arrival data identifying estimated times of arrival for the appointments, and may perform actions based on the estimated time of arrival data.

Abstract: A device may receive sensor data and video data associated with a vehicle, and may process the sensor data, with a rule-based detector model, to determine whether a probability of a vehicle accident satisfies a first threshold. The device may preprocess acceleration data of the sensor data to generate calibrated acceleration data, and may process the calibrated acceleration data, with an anomaly detector model, to determine whether the calibrated acceleration data includes anomalies. The device may filter the sensor data to generate filtered sensor data, and may process the filtered sensor data and anomaly data, with a decision model, to determine whether the probability of the vehicle accident satisfies a second threshold. The device may process the filtered sensor data, the anomaly data, and the video data, with a machine learning model, to determine whether the vehicle accident has occurred, and may perform one or more actions.

Abstract: A device may obtain video data associated with a driving event involving a first vehicle. The device may determine a vanishing point associated with the video data and may construct a cone of impact of the first vehicle based on the vanishing point. The device may detect a second vehicle within the cone of impact and may analyze the subset of video frames to determine a distance between the first vehicle and the second vehicle. The device may determine a speed of the first vehicle during a time period associated with a subset of video frames. The device may determine a headway score, representative of a severity associated with the first vehicle being within a proximity threshold of the second vehicle during the time period, based on the distance and the speed. The device may determine an occurrence of a tailgating event based on the headway score.