Abstract: A method of training a deep neural network, such as would be used in facial recognition, includes training the deep neural network to normalize feature vectors to a learned value representing the radius of a multi-dimensional hypersphere using a convex augmentation of the primary loss function.

Abstract: Disclosed herein is a highly energy-efficient architecture targeting the ultra-low-power sensor domain. The architecture achieves high energy-efficiency while maintaining programmability and generality. The invention introduces vector-dataflow execution, allowing the exploitation of the dataflows in a sequence of vector instructions and to amortize instruction fetch and decode over a whole vector of operations. The vector-dataflow architecture allows the invention to avoid costly vector register file accesses, thereby saving energy.

Abstract: A system and method implementing an acoustic mapping suite that equips intelligent vehicles with the ability to identify and respond to primary sounds of interest in their environment. The invention performs sound separation, localization and classification techniques, while also generating a sound-based map of the vicinity as well as making inferences and qualitative assessments about sound sources with information over time.

Abstract: A strain sensor utilizes an ohmic-based contact switch to detect strain. The sensor can be incorporated into other structures, such as an artificial flapping wing, to detect strain and other parameters, including air flow disturbances. The sensors are fabricated using an additive manufacturing process, with a layer of gold or other conductive material applied for electrical conductivity and UV laser ablation for electrical isolation. The sensor design incorporates mechanical amplification, converting small strains into larger displacements that close contact pads, resulting in an ohmic switch activated at a specific strain threshold. Unlike traditional sensors, the switch provides a high or low state output directly without the need for additional amplification or post-processing. The device can detect disturbances in flapping wing cycles and obtain yaw rotation information, with potential applications in other aircraft for disturbance detection.

Abstract: A hybrid graphene material includes a functional group to expand the use of graphene in various applications. The hybrid material may include a substrate, such as silicon nanowires, where the graphene is fabricated on the surface of the substrate with an out-of-plane topography. Functional groups can be added to the graphene and affect the electrical, chemical, or photo characteristics of the hybrid material.

Abstract: A high gravimetric and volumetric power density heat exchanger is provided for high-temperature and high-pressure applications with counter-flowing hot and cold agents that enter via respective inlet headers, transit parallel and adjacent flow passages, and exit via respective outlet headers. One or more headers for conveying one of the substances may be situated within the flow of the other substance. Structures within the flow passages promote the transfer of heat while limiting pressure drop on one or both sides. The structures may include microscale pins, an array of pins (with specified aspect ratios and spacing), a lattice of interconnected pins, parallel ridges, and/or other features. Through an additive manufacturing process, the headers are monolithically integrated into the heat exchanger instead of being separately constructed and attached.

Abstract: Provided is a method of training a machine-learning-based artificial intelligence (AI) model to handle diverse types of motions occurring during image acquisition, including capturing image data including motion between an imaging device and tissue, modifying the captured image data, resulting in modified image data, by at least one of: altering an amount of time between any two frames; removing a subsequence of frames from the captured image data; and adding a subsequence of one or more new frames to the captured image data, and training a machine-learning-based AI model based on the modified image data. Other systems and methods are also described.

Abstract: An aquatic vessel, illustratively a pontoon boat including a thruster system is disclosed. The aquatic vessel executes a process to automatically position the aquatic vessel relative to a target location such as a mooring implement. Exemplary mooring implements include a dock, a slip, or a lift.

Abstract: Provided are methods for labeling ultrasound data. The method may include training a convolutional neural network (CNN) based on ultrasound data. The ultrasound data may include ultrasonic waveform data (e.g., radio frequency (RF) waveform data). An RF input of each downsampling layer of a plurality of downsampling layers in the CNN may be downsampled. The RF input may include RF waveform data for an ultrasound. Tissues in the ultrasound may be segmented based on an output of the CNN. A system is also disclosed.

Abstract: Disclosed herein is a method to disentangle linear-encoded facial semantics from facial images without external supervision. The method uses linear regression and sparse representation learning concepts to make the disentangled latent representations easily interpreted and manipulated. Generated facial images are decomposed into multiple semantic features and latent representations are extracted to capture interpretable facial semantics. The semantic features may be manipulated to synthesize photorealistic facial images by sampling along vectors representing the semantic features, thereby changing the associate semantics.

Abstract: During training of deep neural networks, a Copernican loss (LC) is designed to augment a primary loss function, for example, a standard Softmax loss, to explicitly minimize intra-class variation and simultaneously maximize inter-class variation. Copernican loss operates using the cosine distance and thereby affects angles leading to a cosine embedding, which removes the disconnect between training and testing.

Abstract: Disclosed herein is an improved method for identifying images containing objects-of-interest from a large set of images. The method comprises mixing two or more of the images to create a grouped image and exposing the grouped image to an object detector trained on grouped images to make an initial determination that the grouped image was formed from at least one image containing an object-of-interest. The images which formed the grouped image are then exposed to regular object detectors to determine a classification of the object-of-interest.

Abstract: Provided is a system, method, and computer program product for segmenting vessels in an ultrasound image. The method includes detecting edges of a vessel in the ultrasound image; detecting a vessel contour of the vessel in the ultrasound image based on the detected edges and a distance regularized level set evolution; and tracking the vessel contour with a Kalman Filter.

Abstract: The Reconfigurable Wheel-Track (RWT) is a novel mechanism that allows a wheel to transform into a track, and vice-versa. The wheel permits a vehicle to travel quickly over smooth and semi-rough terrain, then, on-the-fly, transform rapidly into a powered track for crossing extreme terrain. The reconfigurable wheel-track consists of several main components: an outer tire/tread, drive mechanisms for the wheel and track, support mechanisms for the outer tread when in either wheel mode or track mode, and a reconfiguration mechanism that facilitates the transformation from a wheel to a track and vice-versa. The reconfigurable wheel-track includes sensing, actuation, and controls to facilitate efficient and effective transition from wheel to track and vice-versa, and securely maintain each shape.

Abstract: An automated inventory monitoring system includes an image capture module able to create an image of an aisle of a retail store. Images of products labels are identified in the image and classified as shelf labels or peg labels. For shelf labels, an area of the shelf is defined and associated with the shelf label. Images of products are identified in the image and products on the shelf within an area associated with a shelf label are associated with the shelf label. Products located below a peg label are associated with the peg label. Based on the association between labels and products, out-of-stock products, plugs and spread may be detected and reported to the staff of the retail store.

Abstract: A method of manufacturing a porous part includes controlled freeze casting of a slurry. After freezing, a solvent in the slurry is removed by sublimation and the remaining material is sintered to form the porous part. Spatial and temporal control of thermal conditions at the boundary and inside of the mold can be controlled to create parts with controlled porosity, including size, distribution, and directionality of the pores. Porous parts with near-net-shape from ceramics, metals, polymers and other materials and their combinations can be created.

Abstract: A high-density storage system for goods is described in which totes carrying the goods are storage in a storage structure and stored and retrieved by robotic carriers. The carriers move laterally and/or longitudinally along the exterior of the support structure and retrieve totes from the interior of the structure by manipulating rows of coupled totes. Totes at the ends of rows are quickly removed and stored in another row until the desired tote appears at the end of the row, at which point the carrier proceeds with the tote to the exit point of the storage system. Storing totes is also a quick action by pushing them into any row. As a tote is pushed into the row, it will automatically couple with a tote inside the row that it comes into contact with.

Abstract: Disclosed herein is a system architecture that structures commodity heterogeneous interconnected computing platforms around universal object abstractions, which are a fundamental system abstraction and building block that provides practical and provable end-to-end guarantees of security, correctness, and timeliness for the platform.

Abstract: A system includes an inertial sensing device having an inertial sensor and plurality of stress sensors configured to measure stress applied to the inertial sensing device, and at least one computing device configured to: receive sensor data from the inertial sensor and the plurality of stress sensors; and determine a drift compensation of the inertial sensor based on the sensor data. Other systems, methods, and devices are disclosed.

Abstract: Provided is a system, method, and computer program product for creating a deep-learning model for processing image data. The method includes establishing dense connections between each layer of a plurality of layers of a convolutional neural network (CNN) and a plurality of preceding layers of the CNN, downsampling an input of each downsampling layer of a plurality of downsampling layers in a first branch of the CNN, and upsampling an input of each upsampling layer of a plurality of upsampling layers in a second branch of the CNN by convolving the input.