A5100687474- Advanced Image and Video Retrieval Techniques
- Energetic Materials and Combustion
- Remote-Sensing Image Classification
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Infrastructure Maintenance and Monitoring
- Video Surveillance and Tracking Methods
- Advanced Image Processing Techniques
- Image Enhancement Techniques
- Advanced Graph Neural Networks
- Thermal and Kinetic Analysis
- Automated Road and Building Extraction
- Combustion and Detonation Processes
- Concrete and Cement Materials Research
- Advanced Image Fusion Techniques
- Topic Modeling
- Robotics and Sensor-Based Localization
- Solar Radiation and Photovoltaics
- Asphalt Pavement Performance Evaluation
- Video Analysis and Summarization
- Medical Image Segmentation Techniques
- Atmospheric aerosols and clouds
- Structural Behavior of Reinforced Concrete
- Image Retrieval and Classification Techniques
- Multimodal Machine Learning Applications
University of Electronic Science and Technology of China
2022-2025
Shandong University
2023-2025
Southeast University
2022-2024
Beijing Jiaotong University
2016-2024
University of Delaware
2024
Ministry of Education of the People's Republic of China
2024
Yangzhou University
2024
Beijing Technology and Business University
2024
Sun Yat-sen University
2024
Dalian University of Technology
2024
Abstract Clouds have an enormous influence on the Earth's energy balance, climate, and weather. Cloud types different cloud radiative effects, which is essential indicator of effect radiation. Therefore, identifying type important in meteorology. In this letter, we propose a new convolutional neural network model, called CloudNet, for accurate ground‐based meteorological classification. We build data set, Cirrus Cumulus Stratus Nimbus, consists 11 categories under standards. The total number...
Object detection on the drone faces a great diversity of challenges such as small object inference, background clutter and wide viewpoint. In contrast to traditional problem in computer vision, bird-like angle can not be transplanted directly from common-in-use methods due special texture sky's view. However, lack comprehensive data set, number algorithms that focus using captured by drones is limited. So VisDrone team gathered massive set organized Vision Meets Drones: A Challenge...
Recently, convolution neural network (CNN)-based hyperspectral image (HSI) classification has enjoyed high popularity due to its appealing performance. However, using 2-D or 3-D in a standalone mode may be suboptimal real applications. On the one hand, overlooks spectral information extracting feature maps. other suffers from heavy computation practice and seems perform poorly scenarios having analogous textures along with consecutive bands. To solve these problems, we propose mixed CNN...
Light field image (LFI) quality assessment is becoming more and important, which helps to better guide the acquisition, processing application of immersive media. However, due inherent high dimensional characteristics LFI, LFI turns into a multi-dimensional problem that requires consideration degradation in both spatial angular dimensions. Therefore, we propose novel Tensor oriented No-reference Field Quality evaluator (Tensor-NLFQ) based on tensor theory. Specifically, since regarded as...
Light field image quality assessment (LFI-QA) is a significant and challenging research problem. It helps to better guide light acquisition, processing applications. However, only few objective models have been proposed none of them completely consider intrinsic factors affecting the LFI quality. In this paper, we propose No-Reference Field Quality Assessment (NR-LFQA) scheme, where main idea quantify degradation through evaluating spatial angular consistency. We first measure deterioration...
Cross-modal hashing has received widespread attentions on cross-modal retrieval task due to its superior efficiency and low storage cost. However, most existing methods learn binary codes directly from multimedia data, which cannot fully utilize the semantic knowledge of data. Furthermore, they ranking based similarity relevance data points with multi-label. And usually use a relax constraint hash code causes non-negligible quantization loss in optimization. In this paper, method called Deep...
A large amount of labeled data are important to enhance the performance deep-learning-based methods in area fault diagnosis. Because it is difficult obtain high-quality samples real industrial applications, federated learning an effective framework for solving problem sparse by using distributed data. Its global model updated local client without sharing at each round. Considering computing resources and communication loss multiple clients, efficient method based on stacked autoencoders...
Smart manufacturing, which is increasingly popular worldwide, aided by time-series forecasting. As the volume of historical data increases, powerful forecasting techniques that reveal unknown relationships between past and future values are required to provide accurate forecasts production sales. Thus, in this article, a composite gate recurrent unit (GRU)-Prophet model with an attention mechanism was constructed predict sales volume. In model, Prophet GRU were used capture linear nonlinear...
Sodium aluminosilicate hydrate (NASH) gel is the primary adhesive constituent in environmentally friendly geopolymer. In this study, to understand thermal behavior of material, molecular dynamics was utilized investigate structure, dynamic property, and mechanical NASH subjected temperature elevation from 300 K 1500 K. The skeleton provides plenty oxygen sites accept H-bond invading water molecules. Upon heating, around 18.2% molecules are decomposed produce silicate aluminate hydroxyls....
With the wide use of visual sensors in Internet Things (IoT) past decades, huge amounts images are captured people's daily lives, which poses challenges to traditional deep-learning-based image retrieval frameworks. Most such frameworks need a large amount annotated training data, expensive. Moreover, machines still lack human intelligence, as illustrated by fact that they pay less attention interesting regions humans generally focus on when searching for images. Hence, this paper proposes...
The performance of predicting human fixations in videos has been much enhanced with the help development convolutional neural networks (CNN). In this paper, we propose a novel end-to-end network “SalSAC” for video saliency prediction, which uses CNN-LSTM-Attention as basic architecture and utilizes information from both static dynamic aspects. To better represent each frame, first extract multi-level features same size different layers encoder CNN calculate corresponding attentions, then...
A vaccine is a biological product which an important means for human beings to protect themselves. Most of its users are young children with weak immunity. Once has problem, it will pose serious threat the lives many people. At present, supervision production very simple. The record completely controlled by enterprises. Enterprises only submit records supervisory agency review when needs be sold. Production easily forged and modified. In order solve shortcomings traditional centralized...
Numerous surveillance data processing is crucial in the Internet-of-Things systems with pervasive edge computing. In this process, salient object detection from videos plays an important role because it provides human-concerned semantic cue for various industrial tasks. However, still challenging existing studies two aspects. The first one redundant saliency information moving background to disturb of objects. second difficulty model spatiotemporal uncertainty. To overcome these challenges....