Abstract: Systems and methods for training models to predict dense correspondences across images such as human images. A model may be trained using synthetic training data created from one or more 3D computer models of a subject. In addition, one or more geodesic distances derived from the surfaces of one or more of the 3D models may be used to generate one or more loss values, which may in turn be used in modifying the model's parameters during training.

Abstract: A method for calculating conduction velocity of a cardiac activation wavefront from electrophysiological (“EP”) data points generated by a mapping system during a mapping procedure for a heart is provided. The method comprises for each EP data point comprising a local activation time, and position data defining a location within the heart corresponding to the local activation time: defining a neighborhood of EP points comprising the EP data point and a selection of neighboring EP data points; representing local activation time as a function f(x,y), where x and y our coordinates within a hyperplane defined to contain the neighborhood of EP data points based on the position data for each EP data point in the neighborhood; and calculating conduction velocity at the EP data point as a norm function of a gradient for the function f(x,y).

Abstract: A battery box for an electric vehicle includes an Al casting that extends between and joins to a BIW or chassis rails of the electric vehicle. The Al casting includes a bottom wall, a side wall with a pair of rails that are securely attached to the BIW or chassis rails of the vehicle, and a plurality of cross members. The bottom wall, the side wall, and the plurality of cross members define a structural enclosure with a plurality of battery compartments and at least one crush zone, and the structural enclosure is load bearing member of the electric vehicle.

Abstract: In various examples, a two-dimensional (2D) and three-dimensional (3D) deep neural network (DNN) is implemented to fuse 2D and 3D object detection results for classifying objects. For example, regions of interest (ROIs) and/or bounding shapes corresponding thereto may be determined using one or more region proposal networks (RPNs)—such as an image-based RPN and/or a depth-based RPN. Each ROI may be extended into a frustum in 3D world-space, and a point cloud may be filtered to include only points from within the frustum. The remaining points may be voxelated to generate a volume in 3D world space, and the volume may be applied to a 3D DNN to generate one or more vectors. The one or more vectors, in addition to one or more additional vectors generated using a 2D DNN processing image data, may be applied to a classifier network to generate a classification for an object.

Abstract: According to the invention there is provided a method of improving (e.g. restoring) function in a patient having idiopathic pulmonary fibrosis, which method comprises perorally administering a therapeutically-effective amount of N-butyloxycarbonyl-3-(4-imidazol-1-ylmethylphenyl)-5-iso-butylthiophene-2-sulfonamide (C21), or a pharmaceutically acceptable salt thereof, to said patient. Said treatment is capable of treating said IPF in a therapeutic, including a curative, fashion.

Abstract: A refrigerator includes a filter, a receiver, a manifold, and a cam. The filter has a handle extending therefrom. The receiver defines a cavity configured to receive the filter. The manifold is configured to engage an opposing end of the filter relative to the handle via axial rotation of the filter. The manifold is configured to pivot upward to transition the filter to an activated position. The manifold is configured to pivot downward to transition the filter to a deactivated position. The cam is configured to interfere with the handle in response to partial axial rotation of the filter to prevent the filter from pivoting upward to the activated position. The cam is configured to provide clearance for the handle in response to full axial rotation of the filter to facilitate the filter pivoting upward to the activated position.

Abstract: Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for stochastic candidate selection for content recommendation. An example embodiment operates by a computer-implemented method for stochastic candidate selection for content recommendation. The method includes receiving, by at least one computer processor, a first plurality of content candidates and selecting a second plurality of content candidates from the first plurality of content candidates. The method further include ranking the second plurality of content candidates based on one or more parameters and selecting a third plurality of content candidates from the ranked second plurality of content candidates. The method can further include displaying the third plurality of content candidates using a display device.

Abstract: Implementations described herein identify and correct automatic speech recognition (ASR) misrecognitions. For example, on-device processor(s) of a client device may generate a predicted textual segment that is predicted to correspond to spoken utterance of a user of the client device, and may receive further input that modifies the predicted textual segment to an alternate textual segment. Further, the on-device processor(s) may store these textual segments in on-device storage as a candidate correction pair, and transmit the candidate correction pair to a remote system. Moreover, remote processor(s) of the remote system may determine that the candidate correction pair is an actual correction pair, and may cause client devices to generate updates for a global ASR model for the candidate correction pair. Additionally, the remote processor(s) may distribute the global ASR model to the client devices and/or additional client devices.

Abstract: Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for utilizing a content recommendation system powering a streaming media publisher channel to enhance an ad creative being shown to the user via awareness or performance campaigns. This method allows the platform to present the most relevant hyper-personalized in-channel content to the publisher platform users in endemic banners that run on the platform which then correspondingly helps drive user reach. An example embodiment operates by implementing personalized content banners that may act as a hook for channel users opening their streaming device, both active and lapsed, to enter back into the channel.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing a compressed recurrent neural network (RNN). One of the systems includes a compressed RNN, the compressed RNN comprising a plurality of recurrent layers, wherein each of the recurrent layers has a respective recurrent weight matrix and a respective inter-layer weight matrix, and wherein at least one of recurrent layers is compressed such that a respective recurrent weight matrix of the compressed layer is defined by a first compressed weight matrix and a projection matrix and a respective inter-layer weight matrix of the compressed layer is defined by a second compressed weight matrix and the projection matrix.

Abstract: Methods and systems are described for intelligently managing hero cards generated for a user profile. In an example, a server can collect user interaction data that measures how a user interacts with system components. The system components can include emails, hero cards, and software applications. The server can analyze the user interaction data to determine whether a new hero card type should be enabled for a user profile, whether an active hero card type should be disabled for the user profile, and whether parameters for action options on hero cards should be changed for the user profile. The server can make changes to hero cards for the user profile so that the user can receive customized hero cards based on the user's behavior.

Abstract: A method being implemented via execution of computing instructions configured to run at one or more processors. The method can include determining a primary store and one or more secondary stores for pickup of an order of a user, based at least in part on a pickup type of the order. The method also can include determining real-time availabilities of first time slots at the primary store and real-time availabilities of second time slots at the one or more secondary stores. The method additionally can include generating a list of available time slots comprising at least a portion of the first time slots at the primary store and at least a portion of the second time slots at the one or more secondary stores, based at least in part on the real-time availabilities of the first time slots at the primary store and the real-time availabilities of the second time slots at the one or more secondary stores.

Abstract: Embodiments of the present disclosure provide to techniques for automatically grouping software applications based on their technical patterns/characteristics (i.e., technical facets) via machine learning (ML) algorithms. For instance, a first set of software applications that exhibit a high prevalence of one or more first technical facets may be grouped into a first category, a second set of software applications that exhibit a high prevalence of one or more second technical facets may be grouped into a second category, and so on. Once grouped into categories, the software applications in a given category may be assessed, analyzed, and/or processed together for various purposes.

Abstract: A medical device includes opposing first and second end effectors coupled together to move from an open configuration to a closed configuration, a first link with a distal end pivotally connected to a proximal end of the first end effector, the first link including a first slot at a proximal end of the first link, a second link with a distal end pivotally connected to a proximal end of the second end effector, and a first actuator pin slidable within the first slot.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting utterances of a key phrase in an audio signal. One of the methods includes receiving, by a key phrase spotting system, an audio signal encoding one or more utterances; while continuing to receive the audio signal, generating, by the key phrase spotting system, an attention output using an attention mechanism that is configured to compute the attention output based on a series of encodings generated by an encoder comprising one or more neural network layers; generating, by the key phrase spotting system and using attention output, output that indicates whether the audio signal likely encodes the key phrase; and providing, by the key phrase spotting system, the output that indicates whether the audio signal likely encodes the key phrase.

Abstract: An online concierge system facilitates a checkout bypass process for shoppers. The online concierge system detects when the shopper procuring an order for a customer is ready to exit the warehouse after obtaining the items in the order. The online concierge system applies a trained risk model to automatically determining whether to initiate an audit of the shopper. Responsive to determining to initiate the audit, the online concierge system invokes an auditing process. The online concierge system receives, via an auditor application, verification of the order from the auditor mobile application. Responsive to the verification, the online concierge system completes the order and generates routing instructions via the shopper mobile application for facilitating delivery by the shopper to the customer.

Abstract: Various arrangements for performing transmodulation of a forward feeder link are presented. A first data stream and a second data stream can be modulated into a higher-order modulation forward feeder link having a higher-order digital modulation scheme. A satellite can receive the higher-order modulation forward feeder link. The satellite can demodulate the higher-order modulator forward feeder link into a bit stream. This bit stream may then be remodulated and retransmitted as multiple forward user links.

Abstract: A system for tracking inventory inside a warehouse with put-away accuracy is provided. The system 100 includes unmanned aerial vehicle (UAV) 102 including image capturing device 102A, warehouse 104, inventory tracking unit 106, user device 108, cloud server 110, and network 112. The UAV 102 is configured to capture media contents of pre-defined space within the warehouse 104 using image capturing device 102A. The pre-defined space includes rack bays with unique rack bay identifier and inventory items stocked on rack bays with pallet identifier that is similar to corresponding unique rack bay identifier. The inventory tracking unit 106 determines inventory data including empty space, inventory mismatch, and inventory record by processing the media contents using machine learning models 106A-B. The inventory tracking unit 106 sends empty space alert and mismatch alert to user and transmits inventory data to cloud server 110 and user device 108 through network 112.

Abstract: A cooling plate assembly for dual-sided cooling, the cooling plate assembly can include a first cooling plate and a second cooling plate. The first cooling plate that has a first cooling plate connector side and a first cooling plate cooling side that is opposite the first cooling plate connector side. The second cooling plate that has a second cooling plate connector side and a second cooling plate cooling side that is opposite the second cooling plate connector side. The second cooling plate is configured to attach to the first cooling plate such that formed between the first cooling plate connector side and the second cooling plate connector side is at least one cooling passage through which a cooling fluid is flowed to transfer heat from both the first cooling plate cooling side and the second cooling plate cooling side to the cooling fluid.

Abstract: A method includes receiving, at a search toolbar, a search query from a machine in a network. The machine has an associated machine profile for participating in the network as an entity. The machine profile includes a machine identifier and machine metadata. A query type is determined from the search query. A search context for the machine is determined using a semantic graph of the network. From a set of services for the network, one or more relevant services to respond to the search query are identified based on the query type and the search context. The search query is applied to the one or more relevant services to obtain a set of responses. A set of relevant results for the search query is determined from the set of responses. The set of relevant results is transmitted to the machine.