- Real-Time Systems Scheduling
- Embedded Systems Design Techniques
- Generative Adversarial Networks and Image Synthesis
- Mobile Agent-Based Network Management
- Petri Nets in System Modeling
- Computer Graphics and Visualization Techniques
- Advanced Statistical Methods and Models
- Visual Attention and Saliency Detection
- Hate Speech and Cyberbullying Detection
- Digital Media Forensic Detection
- Software System Performance and Reliability
- Industrial Technology and Control Systems
- Computational Geometry and Mesh Generation
- Gene expression and cancer classification
- Energy Efficient Wireless Sensor Networks
- Image Retrieval and Classification Techniques
- Cell Image Analysis Techniques
- Video Analysis and Summarization
- Statistical Methods and Inference
- Mobile Ad Hoc Networks
- Interconnection Networks and Systems
- Industrial Automation and Control Systems
- Software Engineering Research
- 3D Shape Modeling and Analysis
- Design Education and Practice
Harbin Institute of Technology
2024
State Key Laboratory of Robotics and Systems
2024
Chinese Academy of Sciences
2003-2021
Shenyang Institute of Automation
2002-2021
University of Chinese Academy of Sciences
2019-2021
Alibaba Group (China)
2021
Carnegie Mellon University
1985
A style-based architecture (StyleGAN2) yields outstanding results in data-driven unconditional generative image modeling. This work proposes a Domain-guided Noise-optimization-based Inversion (DNI) method to perform facial manipulation. It works based on an inverse code that includes: 1) novel domain-guided encoder called Image2latent project the StyleGAN2 latent space, which can reconstruct input with high-quality and maintain its semantic meaning well; 2) noise optimization mechanism set...
The machine learning regression model is based on the assumption of normal distribution. In this paper, we mainly study probability distribution and effect convergence values different loss functions model. Based idea robust homogeneous variance model, solved statistical solution two-dimensional problem by using least square method. maximum likelihood estimation parameters probabilistic are obtained order to compare solving two methods, L1 function L2 used for verification. Through...
Due to the variety of applications and their importance, wireless sensor networks (WSNs) would need be connected Internet. Some approaches have been proposed connect existing TCP/IP networks, such as application-level gateways or overlay networks. However, most consume network bandwidth node energy maintain static structure which is neither scalable nor reliable. In this paper, we describe mobile agent based real-time (MBR) mechanism use software (MSA) paradigm design a dynamic...
The key to success in using distributed control system (DCS) is the timely execution of computational tasks which usually reside on different nodes and communicate with one another accomplish a common goal. Therefore, temporal property underlying network important, special-purpose realtime industrial needed. Foundation fieldbus (FF), characterized by explicitly distinguishing periodic aperiodic message, kinds. FF applies centralized media access mechanism (MAC) support MAC message. As for...
There are mainly three limitations of the traditional facial attribute editing techniques: 1) incapability generating an arbitrary image with high-resolution; 2) being unable to generate and edit new images synthesized by computer 3) limited diversity edited images. This paper presents a method for simultaneously. It incorporates high-resolution generator, multi-label classifier, Generalized Linear Model (GLM). Experimental results show that our can images, computer-synthesized perform...
Alongside the prevalence of mobile videos, general public leans towards consuming vertical videos on hand-held devices. To revitalize exposure horizontal contents, we hereby set forth exploration automated horizontal-to-vertical (abbreviated as H2V) video conversion with our proposed H2V framework, accompanied by an accurately annotated H2V-142K dataset. Concretely, framework integrates shot boundary detection, subject selection and multi-object tracking to facilitate subject-preserving...
The study is about citizens opinions on student loans by analyzing Twitter reactions to Bidens loan cancellation project using the machine-driven classification of open-ended response (MDCOR) and found it saved research time, increased efficiency, ensured authenticity objectivity data. After putting data into application, we that five analysis topics appropriate. content can be predicted seeking relevant word for each case. issues related shows mixed impact forgiveness, with some key terms...
A detailed DSPN (deterministic and stochastic Petri net) model performance analysis of WorldFIP MAC sub-layer is constructed in this paper. Firstly, the mechanism introduced. Secondly, accurate modeling by using DSPN. Thirdly, simulation research based on developed, which gets some valuable quantitative results system.
A detailed DSPN (Deterministic and Stochastic Petri Net) model performance analysis of PROFIBUS MAC sub-layer is constructed in this paper. Firstly, the medium access control protocol mechanism introduced. Secondly, accurate models based on single master station system constructed. Thirdly, presented following aspects: maximum throughout network utilization. Finally, simulation research developed, which not only validates but also gets some valuable quantitative results system.
Flexibility of constructing redundant control system is one the main features Foundation Fieldbus (FF). Based on protocol FF, levels component redundancy are discussed in detail, and an instance FF based with analyzed. The analysis indicates that flexible for configuring a which plays important role completely improving reliability system.