A5083390005- Video Surveillance and Tracking Methods
- Cloud Computing and Resource Management
- Topic Modeling
- Energy Load and Power Forecasting
- Advanced Vision and Imaging
- Natural Language Processing Techniques
- Face and Expression Recognition
- Image Enhancement Techniques
- Advanced Image and Video Retrieval Techniques
- Multimodal Machine Learning Applications
- Impact of Light on Environment and Health
- Face recognition and analysis
- Advanced Clustering Algorithms Research
- IoT and Edge/Fog Computing
- Big Data and Business Intelligence
- Infrared Target Detection Methodologies
- Distributed and Parallel Computing Systems
- Data Stream Mining Techniques
- Text and Document Classification Technologies
- Meteorological Phenomena and Simulations
- Infrared Thermography in Medicine
- Wind Turbine Control Systems
- Advanced Memory and Neural Computing
- Human Mobility and Location-Based Analysis
- Software Testing and Debugging Techniques
Dalian University of Technology
2020-2025
The University of Melbourne
2020-2025
Zhejiang Energy Research Institute
2024
Xidian University
2017-2022
Texas Tech University
2017-2021
Hainan University
2020-2021
Central China Normal University
2021
Space Engineering University
2021
The Ohio State University
2017-2020
Baidu (China)
2019
The Visual Object Tracking challenge VOT2019 is the seventh annual tracker benchmarking activity organized by VOT initiative. Results of 81 trackers are presented; many state-of-the-art published at major computer vision conferences or in journals recent years. evaluation included standard and other popular methodologies for short-term tracking analysis as well methodology long-term analysis. was composed five challenges focusing on different domains: (i) VOTST2019 focused RGB, (ii)...
In this study, we propose a novel RGB-T tracking framework by jointly modeling both appearance and motion cues. First, to obtain robust model, develop late fusion method infer the weight maps of RGB thermal (T) modalities. The weights are determined using offline-trained global local multimodal networks, then adopted linearly combine response T Second, when cue is unreliable, comprehensively take cues, i.e., target camera motions, into account make tracker robust. We further switcher switch...
With the popularity of multi-modal sensors, visible-thermal (RGB-T) object tracking is to achieve robust performance and wider application scenarios with guidance objects' temperature information. However, lack paired training samples main bottleneck for unlocking power RGB-T tracking. Since it laborious collect high-quality sequences, recent benchmarks only provide test sequences. In this paper, we construct a large-scale benchmark high diversity UAV (VTUAV), including 500 sequences 1.7...
Unmanned aerial vehicles (UAV) have been widely used in various fields, and their invasion of security privacy has aroused social concern. Several detection tracking systems for UAVs introduced recent years, but most them are based on radio frequency, radar, other media. We assume that the field computer vision is mature enough to detect track invading UAVs. Thus we propose a visible light mode dataset called Dalian University Technology Anti-UAV dataset, DUT short. It contains with total...
ABSTRACT Background Cloud Computing has established itself as an efficient and cost‐effective paradigm for the execution of web‐based applications, scientific workloads, that need elasticity on‐demand scalability capabilities. However, evaluation novel resource provisioning management techniques is a major challenge due to complexity large‐scale data centers. Therefore, simulators are essential tool academic industrial researchers, investigate effectiveness algorithms mechanisms in...
Due to long-distance correlation and powerful pretrained models, transformer-based methods have initiated a breakthrough in visual object tracking performance. Previous works focus on designing effective architectures suited for tracking, but ignore that data augmentation is equally crucial training well-performing model. In this paper, we first explore the impact of general augmentations trackers via systematic experiments, reveal limited effectiveness these common strategies. Motivated by...
This work studies product question answering (PQA) which aims to answer product-related questions based on customer reviews. Most recent PQA approaches adopt end2end semantic matching methodologies, map and answers a latent vector space measure their relevance. Such methods often achieve superior performance but it tends be difficult interpret why. On the other hand, simple keyword-based search exhibit natural interpretability through matched keywords, suffer from lexical gap problem. In...
The use of cloud computing for delivering application services over the Internet has gained rapid traction. Since beginning COVID-19 global pandemic, work from home scheme and increased business presence online have created more demand resources. Many enterprises organizations are expanding their private data centres utilizing hybrid or multi-cloud environments IT infrastructure. Because ever-increasing resources, energy consumption carbon emission become a pressing issue. Renewable sources...
Hourly wind power ramps in ERCOT are studied by applying extreme value theory. Mean excess plot reveals that the tail behavior of large hourly indeed follows a generalized Pareto distribution. The location, shape, and scale parameters distribution then determined using mean least square technique, from which risk measures including α quantile at conditional calculated.
Charge prediction is to automatically predict the judgemental charges for legal cases. To convict a person/unit of charge, case description must contain matching instances constitutive elements (CEs) that charge. This knowledge CEs valuable guide judge in making final decisions. However, it far from fully exploited charge literature. In this paper we propose novel method named Constitutive Elements-guided Prediction (CECP). CECP mimics human's identification process extract potential and...
To detect fundus diseases, for instance, diabetic retinopathy (DR) at an early stage, thereby providing timely intervention and treatment, a new grading method based on convolutional neural network is proposed. First, data cleaning enhancement are conducted to improve the image quality reduce unnecessary interference. Second, conditional generative adversarial with self-attention mechanism named SACGAN proposed augment number of images, addressing problems insufficient imbalanced samples....
Given a question and set of answer candidates, triggering determines whether the candidate contains any correct answers. If yes, it then outputs one. In contrast to existing pipeline methods which first consider individual answers separately make prediction based on threshold, we propose an end-to-end deep neural network framework, is trained by novel group-level objective function that directly optimizes performance. Our penalizes three potential types error allows training framework in...
In recent years, discriminative trackers show its great tracking performance, that is mainly due to the online updating using samples collected during tracking. The model could adapt appearance changes of objects and background well after updating. But these have a serious disadvantage wrong may cause severe degradation. Most training in phase are obtained according result current frame. Wrong will be when inaccurate, seriously affecting discrimination ability model. Besides, partial...
In a sequence, the appearance of both target and background often changes dramatically. Offline-trained models may not handle huge variations well, causing tracking failures. Most discriminative trackers address this issue by introducing an online update scheme, making model dynamically adapt background. Although scheme plays important role in improving tracker's accuracy, it inevitably pollutes with noisy observation samples. It is necessary to reduce risk for better tracking. work, we...
Deep learning based visual trackers entail offline pre-training on large volumes of video datasets with accurate bounding box annotations that are labor-expensive to achieve. We present a new framework facilitate for sequences, which investigates selection-and-refinement strategy automatically improve the preliminary generated by tracking algorithms. A temporal assessment network (T-Assess Net) is proposed able capture coherence target locations and select reliable results measuring their...
An Apache Spark based distributed computing and storage system designed for large-scale health data. The provides the solution data digitalization analysis, while enabling high-throughput processing, real-time processing messaging capabilities. design of has potential to provide many health-related services medical professionals, such as retrieving/processing, alerts, mining. article described key considerations throughout designing process, including comparison different component, finding...
Texture classification is an important research topic in image processing. In 2012, scattering transform computed by iterating over successive wavelet transforms and modulus operators was introduced. This paper presents new approaches for texture features extraction using transform. Scattering statistical cooccurrence are derived from subbands of the decomposition original images. And these used four datasets containing 20, 30, 112, 129 images, respectively. Experimental results show that...
Code retrieval is a key task aiming to match natural and programming languages. In this work, we propose adversarial learning for code retrieval, that regularized by question-description relevance. First, adapt simple technique generate difficult snippets given the input question, which can help of faces bi-modal data-scarce challenges. Second, leverage relevance regularize learning, such generated snippet should contribute more training loss, only if its paired language description...
Frequent Item set Mining (FIM) is a very effective method for knowledge acquisition from data, but with the advent of era big traditional algorithms based on memory are facing severe challenges such as computation speed and storage capacity. Fortunately, Map Reduce model provides an efficient framework distributed programming operation framework. This paper proposes novel Reduce-based H-mine algorithm (MRH-mine), version in environment. Experimental results show that MRH-mine has better...
Compared with other recommendation algorithms, Matrix decomposition is frequently used in the current system. It can not only lead to better results, but also fully take influence of various factors into account, which explains its good scalability. includ-es SVD(Singular Value Decomposition), non-negative matrix decomposition, Latent Factor Model and some traditional techniques designed approximate a high-dimensional low-dimensional. As perfect technique system, SVD traditionally expert at...
The COVID-19 global pandemic is an unprecedented health crisis. Many researchers around the world have produced extensive collection of literature since outbreak. Analysing this information to extract knowledge and provide meaningful insights in a timely manner requires considerable amount computational power. Cloud platforms are designed power on-demand elastic manner. Specifically, hybrid clouds, composed private public data centers, particularly well suited deploy computationally...