A5100770779- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Advanced Image Processing Techniques
- Machine Learning in Materials Science
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
- Gait Recognition and Analysis
- Image Enhancement Techniques
- Blind Source Separation Techniques
- Advanced Vision and Imaging
- Image and Video Quality Assessment
- Physical Activity and Education Research
- Image Processing Techniques and Applications
- Advanced Technologies in Various Fields
- Stellar, planetary, and galactic studies
- Process Optimization and Integration
- Network Traffic and Congestion Control
- Air Quality Monitoring and Forecasting
- Human Pose and Action Recognition
- Wireless Signal Modulation Classification
- Phase Equilibria and Thermodynamics
- Astronomy and Astrophysical Research
- Ionic liquids properties and applications
- Thermodynamic properties of mixtures
- Advanced Battery Materials and Technologies
Huazhong University of Science and Technology
2023-2025
Xi'an Jiaotong University
2022-2023
Peking University
2023
University of Science and Technology of China
2021
Guangxi University
2020
Guangdong University of Technology
2018-2019
Ministry of Education of the People's Republic of China
2017
National University of Defense Technology
2009-2012
Discriminative representation is essential to keep a unique identifier for each target in Multiple object tracking (MOT). Some recent MOT methods extract features of the bounding box region or center point as identity embeddings. However, when targets are occluded, these coarse-grained global representations become unreliable. To this end, we propose exploring diverse fine-grained representation, which describes appearance comprehensively from and local perspectives. This represen-tation...
Occlusion scenarios pose a great challenge to person re-identification (ReID) task because various occlusions may weaken the discriminative features and introduce interference. Recently, Transformer-based networks, which can aggregate of all image patches construct global adaptively, have shown advantages in occluded ReID. Existing methods mainly adopted Transformer as feature extractor enhanced local from output encoder. However, during processing self-attention blocks, disturbing be...
Based on the spectroscopic and shear catalogs for SDSS galaxies in local Universe, we compare optically-selected active galactic nuclei (AGNs) with control star-forming quiescent galactic, inter-halo larger scales. We find that AGNs are preferentially found two specific stages of galaxy evolution: star-burst `green valley' phases, stellar population their host is quite independent mass, different from normal galaxies. Combining galaxy-galaxy lensing clustering large scales, measure mass AGN...
Epilepsy is one of the most common disorders brain. Clinically, to corroborate an epileptic seizure-like symptom and find seizure localization, electroencephalogram (EEG) data are often visually examined by a clinical doctor detect presence epileptiform discharges. Epileptiform discharges transient waveforms lasting for several tens hundreds milliseconds mainly divided into 7 types. It important develop systematic approaches accurately distinguish these from normal control ones. This...
Neural style transfer recently has become one of the most popular topics in academic research and industrial application. The existing methods can generate synthetic images by transferring different styles some to another given content images, but they mainly focus on learning low-level features with losses style, leading greatly alter salient information semantic level. In this paper, an improved scheme is proposed keep regions transferred image same that image. By adding region loss...
As an active microwave coherent imaging technology, synthetic aperture radar (SAR) images suffer from severe speckle noise and low-resolution problems due to the limitations of system, which cause difficulties in image interpretation target detection. However, existing SAR super-resolution (SR) methods usually reconstruct by a determined degradation model hardly consider multiplicative noise, meanwhile, most SR models are trained with datasets (LR) down-sampled their high-resolution (HR)...
Ionic liquids (ILs) are promising electrolytes or solvents for numerous applications owing to their unique properties. However, it is a challenge design the ideal IL with required Variational autoencoders (VAEs) trained by significantly large datasets have shown good performance in drug discovery. low generation efficiency and small sparse prevent application on IL. In this work, we propose high molecular model IL, which realizes multi-objective optimization dataset. The combines VAE,...
The symmetrical difference kernel SAR image edge detection algorithm based on the Canny operator can usually achieve effective of a single view image. When detecting multi-view edge, it has disadvantage low accuracy. An for symmetric nuclear GAN network model is proposed. Multi-view data are generated by model. According to results generation, an arbitrary direction constructed. A non-edge eliminated post-processing. Hough transform used calculate realize accurate experimental show that...
Image colorization is a creative process of reasonably adding colors on gray-scale images to generate well-pleasing colorized images. The existing methods normally require user-supplied hints color points and doodles, or handpicked reference for transferring colors, diverse predicting the results; but final results generated by most them may seem unnatural as consequence unprofessional users' skills, inaccurately transferring, limited scale image collection. To overcome these limitations,...
We present an end-to-end trainable framework for P-frame compression in this paper. A joint motion vector (MV) and residual prediction network MV-Residual is designed to extract the ensembled features of representations information by treating two successive frames as inputs. The prior probability latent modeled a hyperprior autoencoder trained jointly with network. Specially, spatially-displaced convolution applied video frame prediction, which kernel each pixel learned generate predicted...
High-Order Cumulants (HOCs) is widely used as the feature in automatic modulation classification (AMC) for it has outstanding resiliency to noise. However, traditional works require more than Nyquist sampling rate HOCs extraction. In this work, a HOCs-based method based on compressive sensing (CS-HOC) introduced. Without reconstructing original signal, we propose scheme estimate fourth-order and sixth-order cumulants of unknown signals received samples, which greatly reduces number samples....
IPTV transfers network TV signals based on IP protocol. Monitoring the data transmission quality is of great importance to extensive deployment and persistent improvement services. RFC4445 proposes metric MDI (Media Delivery Index) in which DF (Delay Factor)and MLR Loss Rate) are utilized do quantitative evaluation video. However, it quite difficult complicated compute an accurate value process evaluation. To this end, we present TBM (Token Bucket Monitoring) algorithm. Compared with method,...
Discriminative representation is essential to keep a unique identifier for each target in Multiple object tracking (MOT). Some recent MOT methods extract features of the bounding box region or center point as identity embeddings. However, when targets are occluded, these coarse-grained global representations become unreliable. To this end, we propose exploring diverse fine-grained representation, which describes appearance comprehensively from and local perspectives. This requires high...
Zhang, Z.; Liu, Y.; Yang, P.; Ye, W.; Weng, S.; T., and Li, Y., 2018. Real-time rendering technology of large-scale underwater scene on web terminal. In: Z.L. Mi, C. (eds.), Advances in Sustainable Port Ocean Engineering. Journal Coastal Research, Special Issue No. 83, pp. 490–494. Coconut Creek (Florida), ISSN 0749-0208.With the continuous development computer software hardware technology, three-dimensional graphics has gradually become one main research contents graphics. It plays a great...