A5045867920- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Cancer-related molecular mechanisms research
- Advanced Computational Techniques and Applications
- Advanced Text Analysis Techniques
- Advanced Image and Video Retrieval Techniques
- Face and Expression Recognition
- Natural Language Processing Techniques
- Image Retrieval and Classification Techniques
- Advanced Neural Network Applications
- Remote-Sensing Image Classification
- Web Data Mining and Analysis
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
- Human Pose and Action Recognition
- Energetic Materials and Combustion
- Fire Detection and Safety Systems
- Video Surveillance and Tracking Methods
- Semantic Web and Ontologies
- COVID-19 diagnosis using AI
- Image and Signal Denoising Methods
- Semiconductor Lasers and Optical Devices
- Machine Learning and ELM
- Text and Document Classification Technologies
- Rough Sets and Fuzzy Logic
Shenzhen Children's Hospital
2025
Shanghai Clinical Research Center
2025
Harbin Institute of Technology
2009-2024
Shenzhen Institute of Information Technology
2022-2024
Nanjing University of Posts and Telecommunications
2021-2023
Beijing University of Posts and Telecommunications
2006-2023
Beihang University
2023
Shanghai Jiao Tong University
2023
Shenzhen University
2019-2022
China Academy of Space Technology
2014-2022
Unsupervised domain adaptive object detection (UDAOD) aims to learn a detector by generalizing knowledge from labeled source an unlabeled target domain. Though the existing graph-based methods for UDAOD perform well in some cases, they cannot proper node set graph. In addition, these build graph solely based on visual features and do not consider linguistic carried semantic prototypes, e.g., dataset labels. To overcome problems, we propose cross-modality reasoning adaptation (CIGAR) method...
Machine learning models trained in one domain perform poorly the other domains due to existence of shift. Domain adaptation techniques solve this problem by training transferable from label-rich source label-scarce target domain. Unfortunately, a majority existing rely on availability target-domain data, and thus limit their applications small community across few computer vision problems. In paper, we tackle challenging zero-shot (ZSDA) problem, where data is non-available stage. For...
As part of our ongoing search for high-performance nanoenergetic materials, we herein used self-assembly guided by negatively charged graphene oxide (GO) to fabricate a high-energy Al/GO/CuFe2O4 nanocomposite with dense layered structure featuring evenly mixed Al and CuFe2O4 nanoparticles uniformly loaded on the GO surface. Investigation effects equivalence ratio (Φ = 1.00–1.75) showed that heat release achieved at optimal value Φ 1.50 (3175 ± 65 J·g−1) exceeded most known thermites,...
Introduction Methicillin-resistant Staphylococcus aureus (MRSA) exhibits diverse genotypes with varying virulence and resistance profiles, particularly in the context of bloodstream infections (BSI). This study investigates prevalence sasX , mupA qacA/B genes among MRSA isolates from southern China analyzes their genetic relatedness. Methods A polymerase chain reaction (PCR) assay was developed to detect presence gene, which is associated nasal colonization, immune evasion, virulence,...
In most sensor network applications, the information gathered by sensors will be meaningless without location of nodes. Node localization has been a topic active research in recent years. Accurate self-localization capability is highly desirable wireless network. This paper proposes genetic algorithm based (GAL). The proposed adopts two new operators: single-vertex-neighborhood mutation and descend-based arithmetic crossover. Four example problems are used to evaluate performance algorithm....
Following significant developments in technology, alternative devices have been applied fieldwork for animal and plant surveys. Thermal-image acquisition cameras installed on unmanned aerial vehicles (UAVs) used surveys the wilderness. This article demonstrates an example of how UAVs can be high mountainous regions, presenting a case study Sichuan snub-nosed monkey with detection rate 65.19% positive individual identification. It also presents model that prospectively predict population size...
Domain adaptation techniques learn transferable knowledge from a source domain to target and train models that generalize well in the domain. Unfortunately, majority of existing are only applicable scenarios target-domain data task interest is available for training, yet this not often true practice. In general, human beings experts generalization across domains. For example, baby can easily identify bear clipart image after learning category animal photo images. To reduce gap between...
In learning-based image processing a model that is learned in one domain often performs poorly another since the samples originate from different sources and thus have distributions. Domain adaptation techniques alleviate problem of shift by learning transferable knowledge source to target domain. Zero-shot (ZSDA) refers category challenging tasks which no target-domain sample for task interest accessible training. To address this challenge, we propose simple but effective method based on...
Four new isoflavone triglycosides, genistein 7-O-β-d-glucopyranoside-4'-O-[(α-l-rhamnopyranosyl)-(1→2)-β-d-glucopyranoside] (1), 7-O-β-d-glucopyranoside-4'-O-[(β-d-glucopyranosyl)-(1→2)-β-d-glucopyranoside] (2), 7-O-α-l-rhamnopyranoside-4'-O-[(α-l-rhamnopyranosyl)-(1→2)-β-d-glucopyranoside] (3), and 7-O-α-l-rhamnopyranoside-4'-O-[(β-d-glucopyranosyl)-(1→2)-β-d-glucopyranoside] (4), together with nine known compounds, namely, 7-O-β-d-glucopyranoside-4'-O-β-d-glucopyranoside, sophorabioside,...
Unsupervised domain adaptive object detection (UDA-OD) is a challenging task that aims to improve the generalization of detectors across domains. Although existing UDA-OD methods have demonstrated their capabilities, they fail investigate two critical correlations in adaptation procedure, ie., i) correlation between features inside an image and ii) domain-invariant domain-specific To take full advantage these correlations, we propose Cyclic Reconstruction Decoupling Adaptation (CRADA)...
In most sensor network applications, the information gathered by sensors will be meaningless without location of nodes. Node localization has been a topic active research in recent years. Accurate self-localization capability is highly desirable wireless network. A fundamental problem distance-based whether given uniquely localizable or not. Flip ambiguity main that can make not localized. It causes large errors estimates. This paper proposes two-phase algorithm (TPLA) for During first...
Diffuse white matter abnormality (DWMA), or diffuse excessive high signal intensity is observed in 50-80% of very preterm infants at term-equivalent age. It subjectively defined as higher than normal periventricular and subcortical comparison to unmyelinated on T2-weighted MRI images. Despite the well-documented presence DWMA, it remains debatable whether DWMA represents pathological tissue injury a transient developmental phenomenon. Manual tracing exhibits poor reliability reproducibility...