Abstract: Described are systems and methods for monitoring and detecting thermal conditions associated with individual cells of a battery. Embodiments of the present disclosure may monitor, detect, and protect against thermal anomalies that may be experienced at individual cells of a battery, without employing the use of thermal sensors (e.g., a thermocouple, thermistor, etc.). The exemplary systems and methods may be implemented in connection with a battery and/or battery pack having cells arranged in a serial configuration that may be employed in an aerial vehicle, such as an unmanned aerial vehicle (UAV). The described systems and methods detect thermal conditions associated with the individual cells of a battery by modeling each cell as an RC circuit and determining the resistance of each cell based on voltage and current measurements. The determined resistance can approximate and/or be representative of the temperature of the respective cell.

Abstract: Techniques for using scoring algorithms utilizing containers for flexible machine learning inference are described. In some embodiments, a request to host a machine learning (ML) model within a service provider network on behalf of a user is received, the request identifying an endpoint to perform scoring using the ML model. An endpoint is initialized as a container running on a virtual machine based on a container image and used to score data and return a result of said scoring to a user device.

Abstract: Techniques for implementing a dynamically resizable memory region for alternative use in a memory are described. The techniques may include using two concurrent address maps corresponding to two address ranges for a memory represented as an array of memory blocks. The first address range can be mapped to the memory with starting addresses of the memory blocks incrementing sequentially along each row. The second address range can be mapped to the memory with starting addresses of the memory blocks incrementing sequentially along each column. When an access request is received having a target address belonging to the first address range, the target address is provided as the memory address to access the memory. When an access request having a target address belonging to the second address range, the target address is translated by address translation logic into a memory address to access the memory.

Abstract: A lightweight filesystem may be provided for remote storage caching. A filesystem may maintain a persistent cache for a data set stored as data files in immutable data objects in a remote data store. Filesystem metadata may be evaluated to determine whether a portion of a data file is stored in the persistent cache according to an offset and length specified in a request. If in the persistent cache, data obtained from a data block in the persistent cache may be returned. If not in the persistent cache, then the remote data store may be accessed and the data file in the immutable data object read to obtain the portion of the data file.

Abstract: An airframe design may include a bonded frame or assembly, and one or more components that may be removably attached to the bonded frame. The bonded frame may include struts, central bulkheads, a tail section, a plurality of wing sections, and motor mounts that are adhered together using adhesive. The one or more attachable components may include a forward fuselage, motors, propellers, motor pod fairings, stabilizer fins, and landing gear that are attached using fasteners. The bonded frame may reduce the number of parts of the airframe design and may also reduce complexity, cost, and weight, while also increasing stiffness and strength. Further, the various attachable components may facilitate fabrication, assembly, and maintenance of an aerial vehicle having the airframe design.

Abstract: Devices and techniques are generally described for selection of objects in image data using natural language input. In various examples, first image data representing at least a first object and first natural language data may be received. In some examples, first embedding data representing the first natural language data may be generated. Second embedding data representing the first image data may be generated. Relative location data indicating a location of the first object in the first image data relative to at least one other object may be generated. The first embedding data, the second embedding data, and the relative location data may be input into a multi-modal transformer model. The multi-modal transformer model may determine that the first natural language data relates to the first object.

Abstract: An automated, single level, package sortation system may include a conveyance mechanism, a package induct station with a first robotic arm, a first drive unit, a cart building station with a second robotic arm, a cart, and a cart transfer station. Each package and cart of the system may have an associated destination. A package may be transferred by a first robotic arm from a conveyance mechanism to a first drive unit, and then transferred by a second robotic arm from a first drive unit to a cart associated with the same destination as the package. Then, the cart may be transferred via a cart transfer station to the destination associated with the cart and all packages contained therein.

Abstract: A network function virtualization service includes an action implementation layer and an action decisions layer. On a flow of network traffic received at the service, the action implementation layer performs a packet processing action determined at the action decisions layer.

Abstract: Systems and/or methods receive an indication associated with a delivery to an address and determine a route associated with the delivery. The systems and/or methods determine a threshold distance to the address associated with displaying images of an environment along the route. As a device associated with a delivery personnel travels within the threshold, the device displays the image(s) of the environment. Additional image(s) are presented based on arriving at the address. The additional image(s) represent a structure associated with the address. The route is displayed along with the additional image(s) to allow the delivery personnel to efficiently make the delivery.

Abstract: Some natural language command processing systems may handle some commands on a user device rather than sending input to another system for processing. Such a system may include an arbitration component for arbitrating between device and/or system processing. The arbitration component may execute in the system and render a device-specific decision as to whether the device will be able to process the input and/or execute the command, based on information known to the system about the device's capabilities. If the arbitration component predicts that the device will not be able to execute the command, the system may execute the command without waiting for a signal from the device. If the arbitration component predicts that the device will be able to execute the command, the system may halt processing to prevent duplicate execution.

Abstract: Devices and techniques are generally described for anomalous network activity detection. In various examples, first application log data comprising a plurality of computer log actions may be received. A hidden Markov model associated may generate a first hidden state vector, where a first value of the first hidden state vector is associated with anomalous activity. A first timing vector associated with a first account may be determined, wherein the timing vector indicates first time periods of typical activity and second periods of typical inactivity for the first account. A first number of elements of the first hidden state vector that have the first value and which correspond to the second time periods of the first timing vector may be determined. Network access may be disabled for the first account.

Abstract: A point-of-sale (POS) system may use biometric or other contactless techniques to determine tokenization data that attributes a transaction with a particular payment account for settlement. A first payment method specified by a user for use at the POS may be issued by a first bank that does not participate in tokenization. A second payment account associated with a different entity and issued by a second bank that does support tokenization is associated with the first payment method. The POS transaction completes, using a token associated with the second payment account. Another transaction then takes place in which the second bank charges the first bank for the amount of the completed transaction.

Abstract: An object-based data storage service receives a request to store a data object in association with a smart data storage tier. Based at least in part on characteristics of the data object, the object-based data storage service identifies and stores the data object in a first location corresponding to a first data storage tier. The object-based data storage service monitors access to the data object to identify a second set of characteristics of the data object. This second set of characteristics is used to determine that the data object is to be transitioned to a second data storage tier. The object-based data storage service, based at least in part on this determination, stores the data object in a second location corresponding to the second data storage tier.

Abstract: Editing and sharing spreadsheets of a spreadsheet editor system are disclosed. Interactions pertaining to requests of users of the spreadsheet editor system may be treated via natural language queries that are received at an interface of said system. When a first user attempts to share a spreadsheet with a second user, various spreadsheet filtering techniques may cause a new version of the given spreadsheet to be generated for the second user depending on permissions and metadata of one or both users of the editable spreadsheet system. References of the second user may additionally be used to modify a spreadsheet for viewing by the second user.

Abstract: Techniques are described for enabling a service provider to determine the power utilization of electrical lineups and sub-lineups powering physical servers in a data center and place virtual machine instances into the physical servers based on the power utilization and a user-specified preference of a virtual machine instance type.

Abstract: Techniques for a cloud-based workload optimization service to identify customer workloads that are optimized to run on burstable instance types. The techniques include identifying workloads that are successfully running on burstable instance types, and using historical-utilization data for those workloads to train classification models. The optimization service can extract feature data from the historical-utilization data, where the feature data represents utilization characteristics that are indicative of burstable workloads. The feature data is then used to train classification models to receive utilization data for candidate workloads, and determine whether the candidate workloads would be optimized for burstable instance types. The optimization service can then migrate suitable workloads to burstable instance types, and/or provide users with recommendations that their workloads are optimized or suitable for burstable instance types.

Abstract: Systems, devices, and methods are provided for depth guided structure from motion. A system may obtain a plurality of image frames from a digital content item that corresponds to a scene and determine, based at least in part on a correspondence search, a set of 2-D keypoints for the plurality of image frames. A depth estimator may be used to determine a plurality of dense depth map for the plurality of image frames. The set of 2-D keypoints and the plurality of dense depth maps may be used to determine a corresponding set of depth priors. Initialization and/or depth-regularized optimization may be performed using the keypoints and depth priors.

Abstract: This disclosure is directed to, in part, techniques for determining occupancy states of a vehicle over time and adjusting settings associated with the vehicle based at least in part on these occupancy states. For instance, an example vehicle may include an array of sensors that may generate sensor data for identifying an occupancy state of a vehicle. These sensors may include cameras, microphones, in-seat weight sensors, seatbelt sensors, door-latch sensors, and the like, which may be used to generate sensor data indicative of a current occupancy state of a vehicle. An occupancy state may include a number of occupants of the vehicle, the location of these occupants within the vehicle, the identity of these occupants, and/or the like.

Abstract: Implementations for determining content perception by a participant in a media session by analyzing content are described. A content signal, such as an audio signal or a video signal, is received during a media session. A transcript of the content can be determined. Feedback, such as a pre-determined phrase, related to content perception by a participant in the media session is determined using a learning algorithm at the device. Metadata regarding the media session is collected in response to the feedback related to the content perception. The indication of the feedback related to the content perception and the metadata regarding the media session is then sent. In response, one or more adjustments to the media session are received. The one or more adjustments to the media session can be determined based on the indication of the feedback related to the content perception, the metadata regarding the media session, and network performance information.

Abstract: Techniques for generating custom rules are described. For example, a system to receive at least one request to create rules based on a policy and code repository files stored by the storage service; analyze the policy to generate a collection of rule candidates; analyze the code repository files to identify labeled code examples that either conform or do not conform to the rule candidates; receive a selection of the labeled code examples; and synthesize at least one rule that includes a precondition that specifies applicability to the selected labeled code examples and a postcondition that expresses a check to be performed contingent on the precondition being satisfied is at least described.

Abstract: Techniques for performing a cross-platform media content personalization are described.

Abstract: Systems, methods, and computer-readable media are disclosed for automated robotic sortation. In one embodiment, an example system may include a first robotic arm configured to grasp a moving object from a conveyor while the conveyor is in motion, a camera system configured to image objects moving on the conveyor, and a first scanner configured to scan machine-readable codes on objects moving on the conveyor. The system may be configured to sort objects from the conveyor using the first robotic arm based at least in part on an object destination.

Abstract: Techniques for using a machine learning model to determine a proper subset of a multimedia file for a viewer based on their interest without the need to actively control a media player timeline are described.

Abstract: A system can be utilized to retrieve media content and rating schemas, to determine maturity ratings for media content. The media content can be utilized to determine segments of data as building blocks associated with mature content. The building blocks can be mapped to content descriptors and rating levels associated with the rating schemas. The building blocks can be compared the media content to identify portions of the media content that have characteristics represented by the building blocks. The building blocks representing the characteristics in the portions of the media content can be utilized to select content descriptors and rating levels associated with the media content. The selected content descriptor and selected rating levels can be utilized to control how, and/or whether, the media content is made available for output to the consumers.

Abstract: In one example, a method comprises: receiving input codes, wherein the input codes represent a computational dataflow graph; traversing the computational dataflow graph to identify single-entry-single-exit (SESE) subgraphs of the computational dataflow graph, wherein each SESE subgraph has a sequence of nodes comprising a root node and a child node and representing a sequence of element-wise operators, wherein the root node receives a single input tensor, and wherein the child node outputs a single output tensor; determining a merged operator for each SESE subgraph; and generating executable instructions for the computational dataflow graph to be executed by a hardware accelerator having a first execution unit and a second execution unit, wherein the executable instructions comprise first executable instructions for the merged operators targeted at the first execution unit, and second executable instructions for other operators of the computational dataflow graph targeted at the second execution unit.

Abstract: An IoT gateway at a client site uses sensing functions and actuating functions to safely implement actions for equipment. The client provides, to the IoT service, specifications for a sensing function and one or more actuating functions that are to communicate with the sensing function. The functions are deployed to the IoT gateway at the client site. The sensing function may be assigned read only access to the equipment and the actuating function may be assigned write only access to the equipment. The sensing function may collect data from the equipment and determine that a condition has been met (e.g., temperature is above an upper limit). In response, the sensing function provides an indication of an action to the actuating function (e.g., via a unique topic). If the action is a supported action, then the actuating function causes the equipment to perform the action; otherwise, it ignores the indicated action.

Abstract: Systems, methods, and computer-readable media are disclosed for systems and methods for automated generation and presentation of sign language avatars for video content. Example methods may include determining, by one or more computer processors coupled to memory, a first segment of video content, the first segment including a first set of frames, first audio content, and first subtitle data, where the first subtitle data comprises a first word and a second word. Methods may include determining, using a first machine learning model, a first sign gesture associated with the first word, determining first motion data associated with the first sign gesture, and determining first facial expression data. Methods may include generating an avatar configured to perform the first sign gesture using the first motion data, where a facial expression of the avatar while performing the first sign gesture is based on the first facial expression data.

Abstract: Techniques are described for providing logical networking functionality for managed computer networks, such as for virtual computer networks provided on behalf of users or other entities. In some situations, a user may configure or otherwise specify a network topology for a virtual computer network, such as a logical network topology that separates multiple computing nodes of the virtual computer network into multiple logical sub-networks and/or that specifies one or more logical networking devices for the virtual computer network. After a network topology is specified for a virtual computer network, logical networking functionality corresponding to the network topology may be provided in various manners, such as without physically implementing the network topology for the virtual computer network. In some situations, the computing nodes may include virtual machine nodes hosted on one or more physical computing machines or systems, such as by or on behalf of one or more users.

Abstract: Methods, apparatus, and computer-accessible storage media for controlling export of snapshots to external networks in service provider environments. Methods are described that may be used to prevent customers of a service provider from downloading snapshots of volumes, such as boot images created by the service provider or provided by third parties, to which the customer does not have the appropriate rights. A request may be received from a user to access one or more snapshots, for example a request to export the snapshot or a request for a listing of snapshots. For each snapshot, the service provider may determine if the user has rights to the snapshot, for example by checking a manifest for the snapshot to see if entries in the snapshot manifest belong to an account other than the customer's. If the user has rights to the snapshot, the request is granted; otherwise, the request is not granted.

Abstract: Functionality is disclosed herein for using an interactive content service to provide content. A user may interact with the interactive content service to access interactive content, such as interactive fiction, interactive text adventure games, and other interactive content. A user may also use the interactive content to receive auditory output from applications configured to provide text-based content. The interactive content service maintains state information that identifies the current state of the interactive content. For example, before starting a particular interactive adventure game, the state identifies a beginning location within the interactive content. As the user interacts with the game, the interactive content service updates the state of the user within the game.

Abstract: Systems and methods are provided for managing provision of—and access to—data sets among instances of function code executing in an on-demand manner. An API is provided by which functions can store data sets to be shared with other functions, and by which functions can access data sets shared by other functions.

Abstract: Disclosed herein are methods and corresponding apparatus and systems for controlling the configuration of a computer system through locking one or more control registers. In some embodiments, a write-enable controller is configured to permit writing to a control register by a software application when the value of a lock bit has been set to indicate that the control register is unlocked. The control register can be locked by setting the lock bit after the control register has been written to, e.g., as part of a system initialization process that places the computer system into a target configuration. After the control register has been locked, the write-enable controller may prevent further writes to the control register, e.g., a write request from the same application that wrote to the control register earlier or a different software application. The locking of the control register can be maintained until system reset.

Abstract: Methods, systems, and computer-readable media for debugging and profiling of machine learning model training are disclosed. A machine learning analysis system receives data associated with training of a machine learning model. The data was collected by a machine learning training cluster. The machine learning analysis system performs analysis of the data associated with the training of the machine learning model. The machine learning analysis system detects one or more conditions associated with the training of the machine learning model based at least in part on the analysis. The machine learning analysis system generates one or more alarms describing the one or more conditions associated with the training of the machine learning model.

Abstract: To perform a beam search operation on an input tensor using a data processor with native hardware support, the data processor can be programmed with a set of instructions. The set of instructions can include a first machine instruction that operates on the input tensor to obtain N largest values in the input tensor, a second machine instruction that operates on the input tensor to obtain indices corresponding to the N largest values in the input tensor, and a third machine instruction that operates on the input tensor to replace the N largest values in the input tensor with a minimum value.

Abstract: Connection loss may be minimized for performing database query engine changes. A distributed database system may include different instances of the query engine that provide access to a database. When an event to change the version of the query engine is detected, a copy of the database may be created and a new instance of the query engine created. Read-only access to the database may be maintained using the different instances of the query engine while the new instance may be upgraded to the different version of the query engine. Upon successful installation of the different version of the query engine at the new instance, the new instance may be given read-write access to the database using the copy of the database and other database instances may be upgraded to the different version of the query engine.

Abstract: Technologies directed to a directional calibration antenna for calibration of an array antenna of a communication system are described. The communication system includes the array antenna which includes a number of antenna elements. The number of antenna elements are located in an area on a first side of a support structure. A directional antenna is located at a first height above a plane of the array antenna and at a periphery of the area. The directional antenna is pointed towards the array antenna and is located within a near field of the array antenna.

Abstract: A distributed database keeps user-defined functions separate from a query engine by using a frontend. The frontend allows a user-defined function to interact with a proxy application processing interface (API) that is based on an API of the query engine. The frontend sends serialized data to the query engine in order to interact with the API of the query engine. The user-defined function is executed in security environments separate from the frontend and the query engine.

Abstract: The following description is directed to a logic repository service. In one example, a method of a logic repository service can include receiving a first request to generate configuration data for configurable hardware using a specification for application logic of the configurable hardware. The method can include generating the configuration data for the configurable hardware. The configuration data can include data for implementing the application logic. The method can include receiving a second request to download the configuration data to a host server computer comprising the configurable hardware. The method can include transmitting the configuration data to the host server computer in response to the second request so that the configurable hardware is configured with the host logic and the application logic.

Abstract: At a data management service, a child template associated with a parent group of one or more cells of a first data sheet is stored. In response to detecting that a parent cell has been added to the parent group, a particular child data sheet is instantiated automatically in accordance with the template. In response to a change to a value in a first cell of the child data sheet, a corresponding value of a cell of the first data sheet is modified.

Abstract: A representation of an identifier of a packet flow is stored in a particular entry within a particular entry group of a first object maintained at a packet processing service, along with an expiration criterion for information pertaining to the packet flow, including a representation of an action to be performed. The action is performed after it is retrieved from an element identified within a second object based on an entry identifier of the particular entry and a group identifier of the particular group. In response to receiving a packet of another packet flow, respective indications that one or more in-use entries of the particular entry group (whose expiration criteria are met) are available for re-use are stored, without receiving an indication that the corresponding packet flows have terminated.

Abstract: A first encoding representing a set of detected signals is obtained at a sensor-proximity resource of an object recognition application which also includes resources of an analytics service of a provider network. In response to a determination that a cache at the sensor-proximity resource does not include a second encoding which satisfies a similarity criterion with respect to the first encoding, at least a portion of a partition of a spatial index is obtained from another resource selected using an index partition map. A recognition-based action is initiated based on determining that the partition includes an encoding which satisfies the similarity criterion.

Abstract: A database system may include a database service and a separate distributed storage service. The database service (or a database engine head node thereof) may be responsible for query parsing, optimization, and execution, transactionality, and consistency, while the storage service may be responsible for generating data pages from redo log records and for durability of those data pages. For example, in response to a write request directed to a particular data page, the database engine head node may generate a redo log record and send it, but not the data page, to a storage service node. The storage service node may store the redo log record and return a write acknowledgement to the database service prior to applying the redo log record. The server node may apply the redo log record and other redo log records to a previously stored version of the data page to create a current version.

Abstract: A natural language system may be configured to act as a participant in a conversation between two users. The system may determine when a user expression such as speech, a gesture, or the like is directed from one user to the other. The system may processing input data related the expression (such as audio data, input data, language processing result data, conversation context data, etc.) to determine if the system should interject a response to the user-to-user expression. If so, the system may process the input data to determine a response and output it. The system may track that response as part of the data related to the ongoing conversation.

Abstract: Techniques for a computer-implemented service for virtual product placement in video frames are described. According to some embodiments, a computer-implemented method includes receiving, at a virtual product placement service, a request to place a two-dimensional image of a virtual product into a video, identifying, by a machine learning model of the virtual product placement service, a surface depicted in the video for insertion of the two-dimensional image of the virtual product, inserting, by the virtual product placement service, of the two-dimensional image of the virtual product into one or more frames of the video onto the surface to generate a video including the virtual product, and transmitting the video including the virtual product to a viewer device or a storage location.

Abstract: Techniques are generally described for remote estimation of bandwidth. In various examples, a video stream may be received at a first bit rate over a first communication channel. A first value of a network condition of the video stream may be determined over a first time period. A determination may be made that the first value is less than a threshold value. A first bandwidth estimate of the communication channel may be determined. The first bandwidth estimate may comprise the first bit rate reduced by a first percentage. A second value of the network condition may be determined over a second time period. A determination may be made that the second value is greater than the threshold value. A second bandwidth estimate of the communication channel may be determined. The second bandwidth estimate may be less than the first bandwidth estimate.

Abstract: An acoustic event detection system may employ self-supervised federated learning to update encoder and/or classifier machine learning models. In an example operation, an encoder may be pre-trained to extract audio feature data from an audio signal. A decoder may be pre-trained to predict a subsequent portion of audio data (e.g., a subsequent frame of audio data represented by log filterbank energies). The encoder and decoder may be trained using self-supervised learning to improve the decoder's predictions and, by extension, the quality of the audio feature data generated by the encoder. The system may apply federated learning to share encoder updates across user devices. The system may fine-tune the classifier to improve inferences based on the improved audio feature data. The system may distribute classifier updates to the user device(s) to update the on-device classifier.