A5085917066- Advanced Neural Network Applications
- Antenna Design and Analysis
- Domain Adaptation and Few-Shot Learning
- Advanced MIMO Systems Optimization
- Microwave Engineering and Waveguides
- Electric Motor Design and Analysis
- Advanced Manufacturing and Logistics Optimization
- Maritime Ports and Logistics
- Non-Destructive Testing Techniques
- Welding Techniques and Residual Stresses
- Simulation and Modeling Applications
- Optical measurement and interference techniques
- Energy Harvesting in Wireless Networks
- Multimodal Machine Learning Applications
- Ultrasonics and Acoustic Wave Propagation
- Satellite Communication Systems
- Magnetic Bearings and Levitation Dynamics
- Remote-Sensing Image Classification
- Industrial Vision Systems and Defect Detection
- Mobile Agent-Based Network Management
- Advanced Image and Video Retrieval Techniques
- Antenna Design and Optimization
- Video Surveillance and Tracking Methods
- Microfluidic and Capillary Electrophoresis Applications
- Advanced Antenna and Metasurface Technologies
Huazhong University of Science and Technology
2015-2024
Tianjin University
2012-2024
Shanghai Artificial Intelligence Laboratory
2022-2024
China Mobile (China)
2007-2024
Nanjing University of Science and Technology
2024
Beijing Academy of Artificial Intelligence
2022-2024
Chinese Academy of Sciences
2007-2024
Xi'an Jiaotong University
2024
University of Chinese Academy of Sciences
2009-2024
Group Sense (China)
2024
Recently, zero-shot and few-shot learning via Contrastive Vision-Language Pre-training (CLIP) have shown inspirational performance on 2D visual recognition, which learns to match images with their corresponding texts in open-vocabulary settings. However, it remains under explored that whether CLIP, pre-trained by large-scale image-text pairs 2D, can be generalized 3D recognition. In this paper, we identify such a setting is feasible proposing PointCLIP, conducts alignment between...
To quickly detect and count the number of bayberry trees, this paper improves YOLO-v4 model proposes an optimal method for detecting trees based on UAV images. We used Leaky_ReLU activation function to accelerate extraction speed DIoU NMS retain most accurate prediction boxes. In order increase recall rate object detection construct model, K-Means clustering was embedded into NMS. trained using images it determined that threshold 0.25, which had best effect. The a accuracy up 97.78% 98.16%...
Remote sensing image classification (RSIC) is a classical and fundamental task in the intelligent interpretation of remote imagery, which can provide unique labeling information for each acquired image. Thanks to potent global context extraction ability multi-head self-attention (MSA) mechanism, visual transformer (ViT)-based architectures have shown excellent capability natural scene classification. However, order achieve powerful RSIC performance, it insufficient capture spatial alone....
Military-civilian attribute recognition of ships in synthetic aperture radar (SAR) imagery plays an important role marine surveillance. However, high-quality labeled data are hard to obtain for SAR ships, which hinder the development deep learning models. Considering that models directly transferred from optical images cannot achieve satisfactory performance applications due great discrepancy different modalities, we propose a two-stage transfer method by combining data-level and...
Aircraft detection in synthetic aperture radar (SAR) images is still a challenging research task because of the insufficient public data, difficulty multiscale target detection, and complexity background interference. In this article, we construct SAR aircraft dataset (SADD) with complex interference objects to facilitate detection. Then, propose scale expansion feature enhancement pyramid network as SADD baseline. It uses four-scale fusion method combine shallow position information deep...
Currently, under supervised learning, a model pre-trained by large-scale nature scene dataset and then fine-tuned on few specific task labeling data is the paradigm that has dominated knowledge transfer learning. Unfortunately, due to different categories of imaging stiff challenges annotation, there not large enough uniform remote sensing support pre-training in domain (RSD). Moreover, models datasets learning directly fine-tuning diverse downstream tasks seems be crude method, which easily...
In order to improve the resilience of distribution networks after major disasters, a three-level service restoration method considering variability and scarcity generation resources within microgrids is proposed. First, feasible tree from critical loads established maximize amount restored loads. The reserve capacity state incorporated as constraints problem ensure that can have continuous power supply. Then, capability microgrids, secondary scheme using emergency supply vehicle (EPS)...
Object detection with Transformers (DETR) has achieved a competitive performance over traditional detectors, such as Faster R-CNN. However, the potential of DETR remains largely unexplored for more challenging task arbitrary-oriented object problem. We provide first attempt and implement Oriented DEtection TRansformer ($\bf O^2DETR$) based on an end-to-end network. The contributions $\rm O^2DETR$ include: 1) we new insight into oriented detection, by applying Transformer to directly...
Automatic segmentation of lung nodules on computed tomography (CT) images is challenging owing to the variability morphology, location, and intensity. In addition, few methods can capture intra-nodular heterogeneity assist nodule diagnosis. this study, we propose an end-to-end architecture perform fully automated multiple types generate for clinical use. To end, a hybrid loss considered by introducing Faster R-CNN model based generalized intersection over union in generative adversarial...
A 4H SiC betavoltaic nuclear battery is demonstrated. Schottky barrier diode utilized for carrier separation. Under illumination of Ni-63 source with an apparent activity 4 mCi/cm2 open circuit voltage 0.49 V and a short current density 29.44 nA/cm2 are measured. power conversion efficiency 1.2% obtained. The performance the device limited by low shunt resistance, backscattering attenuation electron energy in air electrode. It expected to be significantly improved optimizing design...
In this paper, we present a novel deep metric learning method to tackle the multi-label image classification problem. order better learn correlations among images features, as well labels, attempt explore latent space, where and labels are embedded via two unique neural networks, respectively. To capture relationships between features aim two-way distance over embedding space from different views, i.e., one its is not only smaller than those distances labels' nearest neighbors but also other...
Vehicle detection is a very important application of remote sensing. However, suffering from the low acutance and insufficient color information, weak vehicles in satellite imagery still remains challenge. Image enhancement can improve visual effects sensing images. Nevertheless, most existing image methods aim to quality entire without target guidance, which have ambiguous contributions performance. Methods based on generative adversarial networks (GANs) realized with guidance by addition...
Introduction Canopy stomatal conductance (Sc) indicates the strength of photosynthesis and transpiration plants. In addition, Sc is a physiological indicator that widely employed to detect crop water stress. Unfortunately, existing methods for measuring canopy are time-consuming, laborious, poorly representative. Methods To solve these problems, in this study, we combined multispectral vegetation index (VI) texture features predict values used citrus trees fruit growth period as research...