A5100718116- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Anomaly Detection Techniques and Applications
- Photonic and Optical Devices
- Interstitial Lung Diseases and Idiopathic Pulmonary Fibrosis
- Optical Network Technologies
- Advanced Photonic Communication Systems
- Radiomics and Machine Learning in Medical Imaging
- Vitamin D Research Studies
- Navier-Stokes equation solutions
- Occupational and environmental lung diseases
- Building Energy and Comfort Optimization
- Machine Learning and Data Classification
- Concrete and Cement Materials Research
- Industrial Vision Systems and Defect Detection
- Advanced Computational Techniques and Applications
- Natural Language Processing Techniques
- Human Pose and Action Recognition
- Drilling and Well Engineering
- Topic Modeling
- Oil and Gas Production Techniques
- Reservoir Engineering and Simulation Methods
- Remote-Sensing Image Classification
- Plasma Diagnostics and Applications
Guangzhou University
2025
Beihang University
2021-2024
Donghua University
2024
University of South China
2023-2024
Southeast University
2020-2024
South China Agricultural University
2016-2024
Western University
2024
Shandong Provincial Hospital
2022-2024
Shandong First Medical University
2022-2024
Tsinghua University
2024
Recognizing objects from subcategories with very subtle differences remains a challenging task due to the large intra-class and small inter-class variation. Recent work tackles this problem in weakly-supervised manner: object parts are first detected corresponding part-specific features extracted for fine-grained classification. However, these methods typically treat of each image isolation while neglecting their relationships between different images. In paper, we propose Cross-X learning,...
Convolutional sparse coding (CSC) can model local connections between image content and reduce the code redundancy when compared with patch-based coding. However, CSC needs a complicated optimization procedure to infer codes (i.e., feature maps). In this brief, we proposed convolutional auto-encoder (CSAE), which leverages structure of AE incorporates max-pooling heuristically sparsify maps for learning. Together competition over channels, simple sparsifying strategy makes stochastic...
Resolution is a comprehensive reflection and evaluation index for the visual quality of remote sensing images. Super-resolution processing has been widely applied extracting information from Recently, deep learning methods have found increasing application in super-resolution However, issues such as blurry object edges existing artifacts persist. To overcome these issues, this study proposes an improved generative adversarial network with self-attention texture enhancement (TE-SAGAN) We...
Owing to the competitive advantages of fast response speed, large pushing force, high reliability, and precision, permanent magnet linear synchronous motor (PMLSM) has played an increasingly vital role in various high-speed high-precision control systems. However, PMLSM exhibits nonlinear behavior actual operation, position tracking precision is negatively affected by friction, load changes, other external disturbances. To meet growing demand solve problem for PMLSM, system critical....
We aim to provide a computationally cheap yet effective approach for fine-grained image classification (FGIC) in this letter. Unlike previous methods that rely on complex part localization modules, our learns features by enhancing the semantics of sub-features global feature. Specifically, we first achieve sub-feature semantic arranging feature channels CNN into different groups through channel permutation. Meanwhile, enhance discriminability sub-features, are guided be activated object...
A major progress in deep multilayer neural networks (DNNs) is the invention of various unsupervised pretraining methods to initialize network parameters which lead good prediction accuracy. This paper presents sparseness analysis on hidden unit process. In particular, we use L1 -norm measure and provide some sufficient conditions for that leads with respect popular models-such as denoising autoencoders (DAEs) restricted Boltzmann machines (RBMs). Our experimental results demonstrate when are...
This Letter demonstrates a real-time 100-GbE fiber-wireless seamless integration system operating at the whole W band (75-110 GHz). Based on pair of commercial digital coherent optical modules, transparent transmission 125-Gb/s dual-polarized quadrature phase-shift keying signal has been successfully achieved over two-spans 20-km fiber and up to 150-m electromagnetic single-input single-output wireless link. To best our knowledge, this is first demonstration >100-m distance millimeter-wave...
Mesenchymal stem cells (MSCs) have shown promising therapeutic effects in cell-based therapies and regenerative medicine. Efficient tracking of MSCs is an urgent clinical need that will help us to understand their behavior after transplantation allow adjustment strategies. However, no clinically approved tracers are currently available, which limits the translation cell therapy. In this study, a nanoparticle (NP) for computed tomography (CT)/fluorescence dual-modal imaging,...
Fibroblasts have important roles in the synthesis and remodeling of extracellular matrix (ECM) proteins during pulmonary fibrosis. However, spatiotemporal distribution heterogeneous fibroblasts disease progression remains unknown.In current study, silica was used to generate a mouse model pathological changes lung, single-cell sequencing, spatial transcriptome sequencing an analysis markers cell subtypes were performed identify fibroblast subtypes. A group that play role at early stage...
An advanced Watt class Hall Micro Thruster (HMT) is offered for Space Gravitational Wave Detection (SGWD), and an investigation has been conducted on its start-up, shutdown, rise, fall response times. The research involved creating a circuit thruster discharge signal acquisition, using Faraday probe to measure the plasma plume generated by thruster, characterizing thrust directly. Ultimately time of extracted from current signals. experimental results demonstrate that HMT ranging 1.08 ms...
This paper utilizes a low-power Hall thruster to investigate its flow field's neutral dynamics. Specifically, under free molecular with high Knudsen number (Kn>10), we employ the conductance theory interplay between internal structural parameters of field and conductance, which is directly reflected in gas distribution. Hence, first, increase diameter orifices buffer chamber so that baffle effectively increases thus density (nn) discharge channel increases. phenomenon reduces ionization mean...
We propose a locality-constrained sparse auto-encoder (LSAE) for image classification in this letter. Previous work has shown that the locality is more essential than sparsity task. here introduce concept of into auto-encoder, which enables to encode similar inputs using features. The proposed LSAE can be trained by existing backprop algorithm; no complicated optimization involved. Experiments on CIFAR-10, STL-10 and Caltech-101 datasets validate effectiveness