Abstract: This disclosure relates to detecting visual findings in anatomical images. Methods comprise inputting anatomical images into a neural network to output a feature vector and computing an indication of visual findings being present in the images by a dense layer of the neural network that takes as input the feature vector and outputs an indication of whether each of the visual findings is present in the anatomical images. The neural network is trained on a training dataset including anatomical images, and labels associated with the anatomical images and each of the visual findings. The visual findings may be organised as a hierarchical ontology tree. The neural network may be trained by evaluating the performance of neural networks in detecting the visual findings and a negation pair class which comprises anatomical images where a first visual finding is identified in the absence of a second visual finding.

Abstract: Systems and methods electronically produce a local estimate of less-than-critical resources based on a client side version of digital rules and a coarse values received from an online service provider. Although using the client side version of digital rules and the coarse values may not include all the parameters and values needed to provide a fully accurate estimate of the resource, the ability to locally estimate resources without having to make network calls to the online service provider, such as when there are unfavorable conditions or latency of the network, imminent overloading of the online service provider or other operating conditions or demands on the online service provider preventing it from producing a timely more accurate estimate, provides a faster and more efficient way of obtaining a potentially useful estimate of resources.

Abstract: Washing machine appliances and methods of operating the same are provided. Such methods may include, and/or appliances may be configured for, performing a wash operation on laundry articles received in the wash chamber of the wash basket. After the wash operation, it may be determined that the load of articles is still wet, e.g., that a remaining moisture content of the laundry articles is above a threshold. Based on and in response to determining that the load is still wet, a spin cycle is performed before a beginning of a subsequent wash operation.

Abstract: A system and/or method uses a trained U-Net neural network to remove flow artifacts from optical coherence tomography (OCT) angiography (OCTA) data. The trained U-Net receives as input both OCT structural volume data and OCTA volume data, but expands the OCTA volume data to include depth information. The U-Net applies dynamic pooling along the depth direction and weighs more heavily the portion of the data that follows (e.g., along the contours of) select retinal layers. In this manner the U-Net applies contextually different computations at different axial locations based at least in part on the depth index information and/or (e.g., proximity to) the select retinal layers. The U-net outputs OCT volume data of reduced flow artifacts as compared with the input OCTA data.

Abstract: This disclosure relates to detecting visual findings in anatomical images. Methods comprise inputting anatomical images into a neural network to output a feature vector and computing an indication of visual findings being present in the images by a dense layer of the neural network that takes as input the feature vector and outputs an indication of whether each of the visual findings is present in the anatomical images. The neural network is trained on a training dataset including anatomical images, and labels associated with the anatomical images and each of the visual findings. The visual findings may be organised as a hierarchical ontology tree. The neural network may be trained by evaluating the performance of neural networks in detecting the visual findings and a negation pair class which comprises anatomical images where a first visual finding is identified in the absence of a second visual finding.

Abstract: Techniques and systems are described for generating and using a sound localization model. A described technique includes obtaining for a building a sound sensor map indicating locations of first and second sound sensor devices in respective first and second rooms of the building; causing an autonomous device to navigate to the first room and to emit, during a time window, sound patterns at one or more frequencies within the first room; receiving sound data including first and second sound data respectively from the first and second sound sensor devices that are observed during the time window; and generating and storing a sound localization model based on the sound sensor map, autonomous device location information, and the received sound data, the model being configured to compensate for how sounds travels among rooms in at least a portion of the building such that an origin room of a sound source is identifiable.

Abstract: A method of reducing blood flow within an aneurysm includes: injecting a contrast agent into a blood vessel including an aneurysm; expanding a stent, from a delivery device, across the aneurysm; and confirming that a stagnated area forms in the aneurysm. The stagnated area can form a crescent shape, a mushroom shape, a hemispherical shape, and/or a flat side. Upon confirming that the stagnated area forms in the aneurysm, the delivery device can be withdrawn from the blood vessel. The stagnated area can include the contrast agent. If the stagnated area does not form in the aneurysm, a second occluding device may be deployed. After withdrawing the delivery device, substantially all of the aneurysm progressively thromboses.

Abstract: Systems and methods for facilitating an enterprise user to obtain an answer to a user question within an enterprise based on an enterprise knowledge graph are provided. In particular, an enterprise server may receive the user question from the enterprise user, determine a suggested topic associated with the user question based on the enterprise knowledge graph by transforming the user question into a semantic representation to identify a plurality of similar entities within the enterprise knowledge graph, and determine whether a relevant question-and-answer (Q&A) pair linked to the suggested topic exists based on the enterprise knowledge graph. In response to a determination that the relevant Q&A pair does not exist, the enterprise server may determine a predicted answer to the user question and update the enterprise knowledge graph.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for finding lost objects. In some implementations, a request for a location of an item is obtained. Current video data from one or more cameras is obtained. It is determined that the item is not shown in the current video data. Sensor data corresponding to historical video data is obtained. Events that likely occurred with the item and corresponding likelihoods for each of the events are determined. A likely location for the item is determined based on the likelihoods determined for the events. An indication of the likely location of the item is provided.

Abstract: Systems and techniques for solar energy management are described. A described system includes circuitry to determine a solar power generation value based on a power output of a solar power generator configured to supply electricity to a plurality of devices associated with a property; circuitry to determine a power consumption value of the plurality of devices; and a controller configured to determine a power status based on the solar power generation value and the power consumption value. The controller can be configured to selectively enable additional power consumption among the plurality of devices to an extent of the solar power generation value based on the power status indicating a power surplus state. The controller can be configured to selectively reduce power consumption among the plurality of devices based on the power status indicating a power deficit state.

Abstract: An electronic cash deposit system includes a processor that receives a user-generated request for a deposit of cash into an account, receive data generated from a scan of a machine-readable image on a receptacle that receives the cash, perform a validation of the deposit of the cash based on a correlation between the data generated from the scan and a transaction key, generate confirmation of the deposit of the cash when the validation is confirmed, and send a notification to a banking system to update the account to reflect the deposit of the cash in response to the confirmation being generated.

Abstract: Systems and methods electronically generate sample dots, produce resources associated with the dots and estimate a resource for a target point from known resources of dots near the target point based on client side version of digital rules, cataloged data and coarse values previously received from the online service platform. A client receives cataloged data of a cataloged domain in which the cataloged data includes data representing a plurality of dots and a respective computed resource value for each of the dots and each dot of the plurality of dots represents a point in the cataloged domain. In response to confirming a target point is in the cataloged domain, the system discovers a closest one or more dots to the target point based on the cataloged data, estimates a statistic for a resource for the target point based on the respective computed resource values of the closest one or more dots, stores the estimated statistic in a memory, and produces the local estimate based on the estimated statistic.

Abstract: Embodiments are directed to a system, computer program product, and method for dynamic object mining. A received file is segmented, re-formatted, and organized into buffers, while maintaining the order of the received file. Multiple buffers are processed in parallel for object mining. A listener is encoded into the buffers to support asynchronous processing, and more specifically ordering of mined objects. An output file of the mined objected or associated frames is created. The file is populated with a sequential ordering of the objects that follows the order of the received file.

Abstract: A method of operating a washing appliance includes obtaining images of a wash chamber using a camera assembly while the wash basket is tumbling a load of clothes positioned therein. The one or more images are analyzed using a machine learning image recognition process to determine a tumble condition of the load of clothes in the wash basket, e.g., whether the clothes are tumbling in the wash basket in a manner that generates suitable mechanical washing action. The motor assembly is then used to adjust a basket speed based at least in part on determining that the tumble condition deviates from a target tumble condition.

Abstract: A washing machine appliance may include a tub, a basket, a valve, a spout, a dispenser, a motor, and a controller. The basket may be rotatably mounted and define a chamber for receipt of articles for washing. The valve may be disposed upstream from the tub. The spout may be directed at the tub downstream from the valve. The dispenser may be upstream from the tub for dispensing detergent into the tub. The motor may be in mechanical communication with the basket. The controller may be configured for controlling the washing machine appliance. The controller may be configured to initiate an oversuds-sensitive operation. The oversuds-sensitive operation may include detecting one or more resultant conditions from one or more previous wash cycles of the washing machine appliance, determining an adjusted operating parameter based on the detected resultant conditions, and directing detergent dispensing according to the adjusted operating parameter.

Abstract: A catheter system for infusing a chilled fluid into a subject can include an input lumen, a return lumen, and a transition lumen that provides sufficient flow resistance to urge a first portion of fluid flowing out of the input lumen in the forward direction to flow into the return lumen in the reverse direction toward a cooling device, while a second portion of the fluid flowing out of the input lumen in the forward direction flows out of the distal end of the catheter and into the subject's blood vessel.

Abstract: Features are disclosed for generating intent-specific results in an automatic speech recognition system. The results can be generated by utilizing a decoding graph containing tags that identify portions of the graph corresponding to a given intent. The tags can also identify high-information content slots and low-information carrier phrases for a given intent. The automatic speech recognition system may utilize these tags to provide a semantic representation based on a plurality of different tokens for the content slot portions and low information for the carrier portions. A user can be presented with a user interface containing top intent results with corresponding intent-specific top content slot values.

Abstract: Some example product dispensers that accommodate two full paper towel rolls are provided herein. The product dispenser may employ separate dispensing mechanisms for each product roll and may include roll holders that articulate out from the housing to facilitate easy loading and access to replacement and loading of the other product roll. A roll partition may be utilized to separate the web paths of the two product rolls and enable rotation of front roll holders. Example funnel covers and nip covers can aid with intuitive installation. Example roll holders with retention mechanisms help prevent unintentional dropping of the installed product roll. Other beneficial features include utilizing motor operation sensing to dispense a desired sheet length, automatic/assisted loading into the dispensing mechanism, automatic switching between dispensing of product rolls, and always dispensing from the smaller product roll first. Many other features/structures of example product dispensers are also described herein.