163559 research outputs found
Sort by
Deep Cost Ray Fusion for Sparse Depth Video Completion
No full textIn this paper, we present a learning-based framework for sparse depth video completion. Given a sparse depth map and a color image at a certain viewpoint, our approach makes a cost volume that is constructed on depth hypothesis planes. To effectively fuse sequential cost volumes of the multiple viewpoints for improved depth completion, we introduce a learning-based cost volume fusion framework, namely RayFusion, that effectively leverages the attention mechanism for each pair of overlapped rays in adjacent cost volumes. As a result of leveraging feature statistics accumulated over time, our proposed framework consistently outperforms or rivals state-of-the-art approaches on diverse indoor and outdoor datasets, including the KITTI Depth Completion benchmark, VOID Depth Completion benchmark, and ScanNetV2 dataset, using much fewer network parameters.N
Characterization of meat quality, storage stability, flavor-related compounds, and their relationship in Korean Woorimatdag No. 2 chicken breast meat during cold storage
No full textWe evaluated the quality, storage stability, and flavor-related compounds of breast meat from a novel Korean native chicken breed (Woorimatdag No. 2; WRMD2) and commercial broiler (CB) during seven days of aerobic cold storage. We found that pH and drip loss increased gradually during storage and WRMD2 exhibited a significantly lower pH and higher drip loss than CB. In both groups, aerobic plate counts, volatile basic nitrogen, and lipid oxidation levels increased, whereas creatine and dipeptide levels gradually decreased during storage. WRMD2 exhibited a significantly higher anserine content and lower carnosine-to-anserine ratio than CB. Flavor nucleotide content was influenced more by the storage period, whereas fatty acid composition was affected more by genetic differences. WRMD2 exhibited significantly higher levels of polyunsaturated fatty acids, especially C20:4n6 and C22:6n3, than CB. Interestingly, multivariate analysis highlighted several volatile compounds, including methyl salicylate (day 1), dodecanal (day 3), naphthalene (day 5), and 2,4-decadienal (day 7) as potential biomarkers to distinguish between WRMD2 and CB on each storage day. Correlation analysis identified five key meat quality traits, including drip loss, aerobic plate counts, and anserine that are strongly associated with flavor substances, such as inosine monophosphate, guanosine monophosphate, 1-octen-3-ol, and hexanal. These results offer insights into the distinctive meat quality and flavor compounds in WRMD2 during storage, providing fundamental data that could improve the management and quality of native chicken meat.Y
Text Motion Translator: A Bi-directional Model for Enhanced 3D Human Motion Generation from Open-Vocabulary Descriptions
No full textThe field of 3D human motion generation from natural language descriptions, known as Text2Motion, has gained significant attention for its potential application in industries such as film, gaming, and AR/VR. To tackle a key challenge in Text2Motion, the deficiency of 3D human motions and their corresponding textual descriptions, we built a novel large-scale 3D human motion dataset, LaViMo, extracted from in-the-wild web videos and action recognition datasets. LaViMo is approximately 3.3 times larger and encompasses a much broader range of actions than the largest available 3D motion dataset. We then introduce a novel multi-task framework TMT (Text Motion Translator), aimed at generating faithful 3D human motions from natural language descriptions, especially focusing on complicated actions and those not existing in the training set. In contrast to prior works, TMT is uniquely regularized by multiple tasks, including Text2Motion, Motion2Text, Text2Text, and Motion2Motion. This multi-task regularization significantly bolsters the models robustness and enhances its ability of motion modeling and semantic understanding. Additionally, we devised an augmentation method for the textual descriptions using Large Language Models. This augmentation significantly enhances the models capability to interpret open-vocabulary descriptions while generating motions. The results demonstrate substantial improvements over existing state-of-the-art methods, particularly in handling diverse and novel motion descriptions, laying a strong foundation for future research in the field.N
Artificial Photothermal Nociceptor Using Mott Oscillators
No full textBioinspired sensory systems based on spike neural networks have received considerable attention in resolving high energy consumption and limited bandwidth in current sensory systems. To efficiently produce spike signals upon exposure to external stimuli, compact neuron devices are required for signal detection and their encoding into spikes in a single device. Herein, it is demonstrated that Mott oscillative spike neurons can integrate sensing and ceaseless spike generation in a compact form, which emulates the process of evoking photothermal sensing in the features of biological photothermal nociceptors. Interestingly, frequency-tunable and repetitive spikes are generated above the threshold value (Pth = 84 mW cm-2) as a characteristic of "threshold" in leaky-integrate-and-fire (LIF) neurons; the neuron devices successfully mimic a crucial feature of biological thermal nociceptors, including modulation of frequency coding and startup latency depending on the intensity of photothermal stimuli. Furthermore, Mott spike neurons are self-adapted after sensitization upon exposure to high-intensity electromagnetic radiation, which can replicate allodynia and hyperalgesia in a biological sensory system. Thus, this study presents a unique approach to capturing and encoding environmental source data into spikes, enabling efficient sensing of environmental sources for the application of adaptive sensory systems.Y
Recent Advances in Nanomaterial-Based Biosignal Sensors
No full textRecent research for medical fields, robotics, and wearable electronics aims to utilize biosignal sensors to gather bio-originated information and generate new values such as evaluating user well-being, predicting behavioral patterns, and supporting disease diagnosis and prevention. Notably, most biosignal sensors are designed for body placement to directly acquire signals, and the incorporation of nanomaterials such as metal-based nanoparticles or nanowires, carbon-based or polymer-based nanomaterials-offering stretchability, high surface-to-volume ratio, and tunability for various properties-enhances their adaptability for such applications. This review categorizes nanomaterial-based biosignal sensors into three types and analyzes them: 1) biophysical sensors that detect deformation such as folding, stretching, and even pulse, 2) bioelectric sensors that capture electric signal originating from human body such as heart and nerves, and 3) biochemical sensors that catch signals from bio-originated fluids such as sweat, saliva and blood. Then, limitations and improvements to nanomaterial-based biosignal sensors is depicted. Lastly, it is highlighted on deep learning-based signal processing and human-machine interface applications, which can enhance the potential of biosignal sensors. Through this paper, it is aim to provide an understanding of nanomaterial-based biosignal sensors, outline the current state of the technology, discuss the challenges that be addressed, and suggest directions for development.N
The Use of Telehealth for People With Disabilities: A Systematic Literature Review and Narrative Synthesis
No full textAimsTo identify the use of telehealth for people with disabilities in community or primary care settings and to explore effective telehealth interventions for this group.DesignSystematic literature review and narrative synthesis.Data SourcesThe literature search was conducted in January 2024 using five electronic databases including PubMed, EMBASE, CINAHL, Cochrane library and PsycINFO.MethodsThe review followed the Tawfik's guideline and adhered to the Preferred Reporting Items for Systematic Review and Meta-Analyses guidelines for reporting. Out of 7363 retrieved articles, 1871 duplicates were removed, 5389 were excluded after title and abstract review, and 4 were excluded due to unavailable full text. One additional article was obtained through citation and hand searching. Thirteen studies were quality assessed using the Mixed Methods Appraisal Tool. Quantitative data were narratively synthesised.ResultsThirteen quantitative studies were selected including three quasi-experimental studies and ten randomised controlled trials. The types of telehealth included telemonitoring, computerised intervention, virtual reality, telephone care, mHealth tools, decision support tools, digital storytelling and technology-assisted language interventions. The most common type of disability was intellectual disability, and the most common telehealth provider was the digital device itself. Most studies used surveys as the data collection method and the interventions were mostly conducted individually. Computer-based telehealth interventions demonstrated significant improvement in attention, health knowledge and psychological well-being. Telephone, virtual reality and tablet interventions also had positive impacts on body weight, motor coordination and pragmatic language skills. Telemonitoring was also beneficial.ConclusionsThis systematic review examined the current state and effectiveness of telehealth interventions for people with disabilities. However, few intervention studies were found, and some studies were of poor quality. Continued interest and efforts from the government and researchers are needed targeting people with disabilities.ImpactResults provide valuable insights for healthcare providers, policymakers and researchers. They raise awareness about the potential of telehealth to address healthcare disparities and improve access to care for people with disabilities.Patient or Public ContributionNo patient or public contribution: Systematic review.N
Investigation of different cold adaptation abilities in <i>Salmonella enterica</i> serotype Typhimurium strains using extracellular metabolomic approach
No full textThis study explored the extracellular metabolomic responses of three different Salmonella enterica serotype Typhimurium (S. Typhimurium) strains-ATCC 13311 (STy1), NCCP 16964 (STy4), and NCCP 16958 (STy8)-cultured at refrigeration temperatures. The objective was to identify the survival mechanisms of S. Typhimurium under cold stress by analyzing variations in their metabolomic profiles. Qualitative and quantitative assessments identified significant metabolite alterations on day 6, marking a critical inflection point. Key metabolites such as trehalose, proline, glycerol, and tryptophan were notably upregulated in response to cold stress. Through multivariate analyses, the strains were distinguished using three metabolites-4-aminobutyrate, ethanol, and uridine-as potential biomarkers, underscoring distinct metabolic responses to refrigeration. Specifically, STy1 exhibited unique adaptive capabilities through enhanced metabolism of betaine and 4-aminobutyrate. These findings highlight the variability in adaptive strategies among S. Typhimurium strains, suggesting that certain strains may possess more robust metabolic pathways for enhancing survival in refrigerated conditions.N
Heat-Up Process: Road to Synthesizing Monodisperse Nanoparticles
No full textLarge-scale synthesis of monodisperse nanoparticles is highly desirable for practical applications of nanoparticles in various fields of emerging technology. Among colloidal synthetic routes of monodisperse nanoparticles, heat-up process, which involves a gradual heating of precursor solution in a batch reactor, has received utmost interest after its successful size-controlled synthesis of various kinds of nanoparticles. In this essay, we discuss the fundamental research regarding the synthesis of monodisperse nanoparticles and describe how researchers developed the heat-up process. The mechanistic insights into this synthetic method and the discovery of size-dependent properties of monodisperse nanoparticles are also presented. We further emphasize its great impacts on utilizing monodisperse nanoparticles synthesized via the heat-up process for biomedical technology, energy conversion and storage devices, as well as electronic and optoelectronic devices.N
Miniature Li<SUP>+</SUP> solvation by symmetric molecular design for practical and safe Li-metal batteries
No full textDeveloping high-safety Li-metal batteries (LMBs) with rapid rechargeability represents a crucial avenue for the widespread adoption of electrochemical energy storage devices. Realization of LMBs requires an electrolyte that combines non-flammability with high electrochemical stability. Although current electrolyte technologies have enhanced LMB cyclability, rational electrolyte fabrication capable of simultaneously addressing high-rate performance and safety remains a grand challenge. Here we report an electrolyte design concept to enable practical, safe and fast-cycling LMBs. We created miniature anion-Li+ solvation structures by introducing symmetric organic salts into various electrolyte solvents. These structures exhibit a high ionic conductivity, low desolvation barrier and interface stabilization. Our electrolyte design enables stable, fast cycling of practical LMBs with high stability (LiNi0.8Co0.1Mn0.1O2 cell (twice-excessed Li): 400 cycles) and high power density (pouch cell: 639.5 W kg-1). Furthermore, the Li-metal pouch cell survived nail penetration, revealing its high safety. Our electrolyte design offers a viable approach for safe, fast-cycling LMBs.N
In-sensor multilevel image adjustment for high-clarity contour extraction using adjustable synaptic phototransistors
No full textRobotic vision has traditionally relied on high-performance yet resource-intensive computing solutions, which necessitate high-throughput data transmission from vision sensors to remote computing servers, sacrificing energy efficiency and processing speed. A promising solution is data compaction through contour extraction, visualizing only the outlines of objects while eliminating superfluous backgrounds. Here, we introduce an in-sensor multilevel image adjustment method using adjustable synaptic phototransistors, enabling the capture of well-defined images with optimal brightness and contrast suitable for achieving high-clarity contour extraction. This is enabled by emulating dopamine-mediated neuronal excitability regulation mechanisms. Electrostatic gating effect either facilitates or inhibits time-dependent photocurrent accumulation, adjusting photo-responses to varying lighting conditions. Through excitatory and inhibitory modes, the adjustable synaptic phototransistor enhances visibility of dim and bright regions, respectively, facilitating distinct contour extraction and high-accuracy semantic segmentation. Evaluations using road images demonstrate improvement of both object detection accuracy and intersection over union, and compression of data volume.N