A5100318755- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Medical Image Segmentation Techniques
- Video Surveillance and Tracking Methods
- 3D Shape Modeling and Analysis
- Human Pose and Action Recognition
- Data Management and Algorithms
- Simulation Techniques and Applications
- Video Analysis and Summarization
- Domain Adaptation and Few-Shot Learning
- Higher Education and Teaching Methods
- Topic Modeling
- Innovative Educational Techniques
- Generative Adversarial Networks and Image Synthesis
- Image and Object Detection Techniques
- Visual Attention and Saliency Detection
- Advanced Vision and Imaging
- Face recognition and analysis
- Model-Driven Software Engineering Techniques
- Textile materials and evaluations
- Vehicle License Plate Recognition
- Advanced Database Systems and Queries
- Optimization and Variational Analysis
Manuel L. Quezon University
2024
Alibaba Group (China)
2018-2023
Dalian Maritime University
2023
Alibaba Group (United States)
2019-2022
Union Hospital
2018-2020
Fujian Medical University
2018-2020
Shanghai Municipal Education Commission
2018
Shanghai Jiao Tong University
2018
National University of Defense Technology
2013-2015
Southwest Petroleum University
2015
The task of Human-Object Interaction (HOI) detection could be divided into two core problems, i.e., human-object association and interaction understanding. In this paper, we reveal address the disadvantages conventional query-driven HOI detectors from aspects. For association, previous two-branch methods suffer complex costly post-matching, while single-branch ignore features distinction in different tasks. We propose Guided-Embedding Network (GEN) to attain a pipeline without post-matching....
In 3D face reconstruction, orthogonal projection has been widely employed to substitute perspective simplify the fitting process. This approximation performs well when distance between camera and is far enough. However, in some scenarios that very close or moving along axis, methods suffer from inaccurate reconstruction unstable temporal due distortion under projection. this paper, we aim address problem of single-image Specifically, a deep neural network, Perspective Network (PerspNet),...
The development of online economics arouses the demand generating images models on product clothes, to display new clothes and promote sales. However, expensive proprietary model challenge existing image virtual try-on methods in this scenario, as most them need be trained considerable amounts accompanied with paired images. In paper, we propose a cheap yet scalable weakly-supervised method called Deep Generative Projection (DGP) address specific scenario. Lying heart proposed is imitate...
To improve the rate-distortion (R-D) quality, x265 rate-control makes a variety of vital decisions-such as scene cut detection, slice type decision, and coding-unit quantization parameter (QP) offsets-leveraging on lookahead to evaluate information propagation through current near future consecutive frames. However, frame base QP$ that dominates bit amount allocated one was only determined by long-term complexity history in original algorithm, allocation became insensitive recent changes...
Domain specific modelling, as a widely accepted software development paradigm in engineering community, has attracted lot of attention M&S community for it raises the abstraction level and enables modelling with domain concepts. However, current DSM research is not systematic deep enough to provide generic support simulation systems development, especially complex systems. To fulfill full potential R&D, we need combine fruits from both field field. In this paper, We concentrate on using...
This paper introduces our submission to the 2nd 3DFAW Challenge. To get a high-accuracy 3D dense face shape based on 2D videos or multiple images, framework which is consist of multi-reconstruction branches and mesh retrieval module, proposed effectively utilize information all frames results predicted by branches. The recent state-of-the-art methods single-view multi-view are introduced form an ensemble independent regression networks. candidate each branch synthesized weighted linear...
The rate control of x265 capitalizes on the low-resolution frame pre-motion-estimation to analyze information prorogation in consecutive frames, with which detect scene cut, decide slice type, and adjust quantization ( Q) offsets coding-unit (CU)-level, respectively. In this paper, we increase accuracy frame-level base Q calculation type decision x265. We implemented proposed algorithms version 2.4. Experiments revealed that, 0.5904dB BDPNSR average up 1.7205dB coding quality improvement...
Separating the dominant person from complex background is significant to human-related research and photo-editing based applications. Existing segmentation algorithms are either too general separate region accurately, or not capable of achieving real-time speed. In this paper, we introduce multi-domain learning framework into a novel baseline model construct Multi-domain TriSeNet Networks for single image segmentation. We first divide training data different subdomains on characteristics...
Complex systems contain hierarchical heterogeneous subsystems and diverse domain behavior patterns, which bring a grand challenge for simulation modeling. To cope with this challenge, the M&S community extends their existing modeling paradigms to promote reusability, interoperability composability of models systems; however, these efforts are relatively isolated limited own technical space. In paper, we propose specific (DSM)-based multi-paradigm approach utilizes model driven engineering...
An important feature to be considered in the design of a multimedia database system (MMDBS) is content based retrieval images. Spatial features represent spatial relationships among objects an image. The salient (interesting objects) can organized object hierarchy, on oriented concepts. paper proposes indexing scheme, called 2D-h trees, for This scheme organizes representations images and hierarchical efficiently query optimization. Our performance analysis indicates that 2D-h-tree efficient index
In the era of social media video platforms, popular ``hot-comments'' play a crucial role in attracting user impressions short-form videos, making them vital for marketing and branding purpose. However, existing research predominantly focuses on generating descriptive comments or ``danmaku'' English, offering immediate reactions to specific moments. Addressing this gap, our study introduces \textsc{HotVCom}, largest Chinese hot-comment dataset, comprising 94k diverse videos 137 million...