A5100407437- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Metaheuristic Optimization Algorithms Research
- Video Analysis and Summarization
- Anomaly Detection Techniques and Applications
- Advanced Neural Network Applications
- Organic Electronics and Photovoltaics
- Advanced Image Fusion Techniques
- Remote-Sensing Image Classification
- Infrared Target Detection Methodologies
- Domain Adaptation and Few-Shot Learning
- Organic Light-Emitting Diodes Research
- Advanced Multi-Objective Optimization Algorithms
- Luminescence and Fluorescent Materials
- Machine Learning and Data Classification
- Video Surveillance and Tracking Methods
- Text and Document Classification Technologies
- Network Security and Intrusion Detection
- Advanced Data Compression Techniques
- Image Processing Techniques and Applications
- Machine Learning and ELM
- Evolutionary Algorithms and Applications
- Conducting polymers and applications
- Internet Traffic Analysis and Secure E-voting
- Advanced Vision and Imaging
Xiangnan University
2018-2024
Henan Normal University
2017-2024
National University of Defense Technology
2015-2024
Dalian University of Technology
2016-2024
Columbia University
2013-2024
University of Science and Technology of China
2017-2024
Jilin Business and Technology College
2023
Jilin Province Science and Technology Department
2016-2023
Chinese Academy of Engineering
2015-2023
Guangdong University of Technology
2022
Social media sharing web sites like Flickr allow users to annotate images with free tags, which significantly facilitate Web image search and organization. However, the tags associated an generally are in a random order without any importance or relevance information, limits effectiveness of these other applications. In this paper, we propose tag ranking scheme, aiming automatically rank given according their content. We first estimate initial scores for based on probability density...
High-resolution representations are essential for position-sensitive vision problems, such as human pose estimation, semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image a low-resolution representation through subnetwork that is formed by connecting high-to-low resolution convolutions \emph{in series} (e.g., ResNet, VGGNet), then recover high-resolution from encoded representation. Instead, our proposed network, named High-Resolution...
We presented a novel and simple approach towards white photoluminescence electroluminescence based on blue emissive materials through controlled acid protonation; this methodology has potential applicability in fabricating OLEDs considering the simplicity of both material design device fabrication.
Single-phase "greener" phosphor with broadband emission made of colloidal quantum dots (QDs) is most desirable for use as color convertors in white light-emitting diodes (WLEDs). Here, we demonstrate a single "cadmium-free" component system consisting Cu-doped InP core/ZnS barrier/InP well/ZnS shell QDs, which exhibits two peaks by controlling the barrier thickness under single-excitation wavelength, one being attributed to InP, and other resulting from well. The dual emissive were...
A novel organosilane compound Si(PPI)<sub>2</sub> with good carrier-transport ability has been used as an efficient host for high-performance fluorescent pure-blue and phosphorescent green/red OLEDs.
Recently, learning based hashing methods have become popular for indexing large-scale media data. Hashing map high-dimensional features to compact binary codes that are efficient match and robust in preserving original similarity. However, most of the existing treat videos as a simple aggregation independent frames index each video through combining indexes frames. The structure information videos, e.g., discriminative local visual commonality temporal consistency, is often neglected design...
Analysis and detection of complex events in videos require a semantic representation the video content. Existing methods typically users to pre-define an exhaustive concept lexicon manually annotate presence concepts each video, which is infeasible for real-world event problems. In this paper, we propose automatic discovery scheme by exploiting Internet images their associated tags. Given target its textual descriptions, crawl collection tags performing text based image search using noun...
Two novel fluorescent materials containing phenanthroimidazole/diphenylphosphine oxide possessing bipolar charge-transporting property have been used not only as emitters to fabricate high-performance deep-blue OLEDs, but also hosts highly efficient green and red PhOLEDs.
The popular unsupervised video object segmentation methods fuse the RGB frame and optical flow via a two-stream network. However, they cannot handle distracting noises in each input modality, which may vastly deteriorate model performance. We propose to establish correspondence between modalities while suppressing signals optimal structural matching. Given frame, we extract dense local features from image flow, treat them as two complex structured representations. Wasserstein distance is...
Late fusion addresses the problem of combining prediction scores multiple classifiers, in which each score is predicted by a classifier trained with specific feature. However, existing methods generally use fixed weight for all classifier, and thus fail to optimally determine individual samples. In this paper, we propose sample-specific late method address issue. Specifically, cast into an information propagation process propagates weights learned on labeled samples unlabeled samples, while...
It is known that two-dimensional two-component fundamental solitons of the semivortex (SV) type, with vorticities (s_{+},s_{-})=(0,1) in their components, are stable ground states (GSs) spin-orbit-coupled (SOC) binary Bose-Einstein condensate contact self-attraction acting both spite possibility critical collapse system. However, excited (ESs) SV solitons, vorticity set (s_{+},s_{-})=(S_{+},S_{+}+1) and S_{+}=1,2,3,..., unstable same We construct ESs SOC system opposite signs...
Accurate network traffic identification is an important basis for monitoring and data analysis, the key to improve quality of user service. In this paper, through analysis two methods based on machine learning deep packet inspection, a method inspection proposed. This uses technology identify most traffic, reduces workload that needs be identified by method, can specific application improves accuracy identification. Machine used assist in identifying with encryption unknown features, which...
In this work, three security models of quantum key distribution are introduced. The practical characters the light source, modulator and single photon detector analyzed, corresponding differences between perfect system estimated. details attacking methods measures de- scribed investigated. We also focus on device-independent class protocol, which is a new important aspect research area.
We study the problem of learning with label proportions in which training data is provided groups and only proportion each class group known. propose a new method called proportion-SVM, or $\propto$SVM, explicitly models latent unknown instance labels together known large-margin framework. Unlike existing works, our approach avoids making restrictive assumptions about data. The $\propto$SVM model leads to non-convex integer programming problem. In order solve it efficiently, we two...