A5112399362- Video Coding and Compression Technologies
- Anomaly Detection Techniques and Applications
- Advanced Neural Network Applications
- Advanced Vision and Imaging
- Advanced Data Compression Techniques
- Industrial Vision Systems and Defect Detection
- Image and Video Quality Assessment
- Video Surveillance and Tracking Methods
- Multimodal Machine Learning Applications
- Technology and Security Systems
- Adversarial Robustness in Machine Learning
- Infrastructure Maintenance and Monitoring
- Advanced Image and Video Retrieval Techniques
- Innovative concrete reinforcement materials
- Advanced Image Processing Techniques
- Remote Sensing and LiDAR Applications
- Nonlinear Optical Materials Studies
- Advanced Computational Techniques and Applications
- Isotope Analysis in Ecology
- 3D Shape Modeling and Analysis
- Innovative Microfluidic and Catalytic Techniques Innovation
- Image Processing Techniques and Applications
- Stability and Control of Uncertain Systems
- Complex Network Analysis Techniques
- Stock Market Forecasting Methods
Guangxi University of Science and Technology
2024
Zhejiang University
2024
Shenyang Jianzhu University
2022-2024
Central South University
2024
Huzhou University
2024
Xidian University
2023
Laboratoire Traitement et Communication de l’Information
2023
Télécom Paris
2023
Zhengzhou University of Aeronautics
2022
Guangdong University of Finance
2022
In real-world recommendation scenarios, users engage with items through various types of behaviors. Leveraging diversified user behavior information for learning can enhance the target behaviors (e.g., buy), as demonstrated by recent multi-behavior methods. The mainstream framework consists two steps: fusion and prediction. Recent approaches utilize graph neural networks employ multi-task paradigms joint optimization in prediction step, achieving significant success. However, these methods...
Fine-grained image classification, which aims to distinguish images with subtle distinctions, is a challenging task for two main reasons: lack of sufficient training data every class and difficulty in learning discriminative features representation. In this paper, address the issues, we propose two-phase framework recognizing from unseen fine-grained classes, i.e., zero-shot classification. first feature phase, finetune deep convolutional neural networks using hierarchical semantic structure...
Computer-aided diagnosis using deep learning algorithms has been initially applied in the field of mammography, but there is no large-scale clinical application.This study proposed to develop and verify an artificial intelligence model based on mammography. Firstly, mammograms retrospectively collected from six centers were randomized a training dataset validation for establishing model. Secondly, was tested by comparing 12 radiologists' performance with without it. Finally, prospectively...
This paper addresses the problem of stability analysis and state feedback fuzzy control design for multi-line re-entrant manufacturing systems (RMSs) with impulse-type jumps. Under impulse effect, RMS is modeled as a hyperbolic impulsive partial differential equation (IPDE) that possesses hybrid characteristics. Then, matrix inequality condition depending on discrete dynamic information presented to deal jump behavior IPDE at instants. Moreover, robust controller based parallel distributed...
The accurate detection of tunnel lining cracks and prompt identification their primary causes are critical for maintaining availability. advancement deep learning, particularly in the domain convolutional neural network (CNN) image segmentation, has made crack more feasible. However, CNN-based technique commonly prioritizes increasing algorithmic complexity to enhance accuracy, posing a challenge balancing accuracy efficiency algorithm. Motivated by superior performance Unet this paper...
When dealing with high-dimensional data, such as in biometric, e-commerce, or industrial applications, it is extremely hard to capture the abnormalities full space due curse of dimensionality. Furthermore, becoming increasingly complicated but essential provide interpretations for outlier detection results a consequence large number features. To alleviate these issues, we propose new model based on Variational AutoEncoder and Genetic Algorithm (VAEGA) detecting outliers subspaces data. The...
We propose a new method for modeling the indoor scene from single color image. With our system, user only needs to drag few semantic bounding boxes surrounding objects of interest. Our system then automatically finds most similar 3D models ShapeNet model repository and aligns them with corresponding To achieve this, each is represented as group view-dependent representations generated set synthesized views. iteratively conduct object segmentation retrieval, based on observation that good...
With the in-depth development of grid intelligence, data in power system becomes more and important. However, environment high magnetoelectric interference system, collection transmission are missing due to equipment failure communication interference. In this case, it is necessary complete processing. The traditional reconstruction method ignores inherent relationship between which reduces accuracy. This paper proposes a completion based on Generative Adversarial Imputation Nets. article...
Colored Point Cloud (CPC) is often distorted in the processes of its acquisition, processing, and compression, so reliable quality assessment metrics are required to estimate perception distortion CPC. We propose a Full-reference Quality Metric for colored point cloud based on Graph signal features Color (FQM-GC). For geometric distortion, normal coordinate information sub-clouds divided via segmentation used construct their underlying graphs, then, structure extracted. color corresponding...
First responders (FRs) navigate hazardous, unfamiliar environments in the field (e.g., mass-casualty incidents), making life-changing decisions a split second. AR head-mounted displays (HMDs) have shown promise supporting them due to its capability of recognizing and augmenting challenging hands-free manner. However, design space not been thoroughly explored by involving various FRs who serve different roles firefighters, law enforcement) but collaborate closely field. We interviewed 26...
Self-supervised learning (SSL) shows impressive performance in many tasks lacking sufficient labels. In this paper, we study SSL time series anomaly detection (TSAD) by incorporating the characteristics of data. Specifically, build an algorithm consisting global pattern and local association learning. The module builds encoder decoder to reconstruct raw data detect anomalies. To complement limitation that ignores associations between points their adjacent windows, design a module, which...
Video Object Segmentation (VOS) is a vital task in computer vision, focusing on distinguishing foreground objects from the background across video frames. Our work draws inspiration Cutie model, and we investigate effects of object memory, total number memory frames, input resolution segmentation performance. This report validates effectiveness our inference method coMplex SEgmentation (MOSE) dataset, which features complex occlusions. experimental results demonstrate that approach achieves...
In order to solve the problems of low tensile strength composite mortar prone cracking when reinforced concrete beams are strengthened by traditional methods, this paper proposes a new polyurethane concrete–prestressing wire (PUC–PSW) reinforcement method using (PUC) as embedding material. Twelve T-beams were tested for PUC–PSW flexural reinforcement. These consisted one unreinforced beam, four PSW-reinforced and seven PUC–PSW-reinforced beams. The material, anchorage form, PUC material...
Video Object Segmentation (VOS) presents several challenges, including object occlusion and fragmentation, the dis-appearance re-appearance of objects, tracking specific objects within crowded scenes. In this work, we combine strengths state-of-the-art (SOTA) models SAM2 Cutie to address these challenges. Additionally, explore impact various hyperparameters on video instance segmentation performance. Our approach achieves a J\&F score 0.7952 in testing phase LSVOS challenge VOS track,...
Despite the promising performance of current video segmentation models on existing benchmarks, these still struggle with complex scenes. In this paper, we introduce 6th Large-scale Video Object Segmentation (LSVOS) challenge in conjunction ECCV 2024 workshop. This year's includes two tasks: (VOS) and Referring (RVOS). year, replace classic YouTube-VOS YouTube-RVOS benchmark latest datasets MOSE, LVOS, MeViS to assess VOS under more challenging environments. attracted 129 registered teams...
A wood species classification scheme was developed based on open set using an improved Nearest Non-Outlier (NNO) classifier. Near infrared (NIR) spectral curves were collected in band 950 to 1650 nm by a micro spectrometer. The dimension reduction performed with Metric Learning (ML) algorithm. Two improvements proposed the following NNO First, cluster analysis each class Density Peak Clustering (DPC) algorithm get 1 3 clusters. fixed threshold for all classes replaced variable This defines...
The traditional video broadcast system can not adapt to the variable terminals and heterogeneous network in modern times, this paper presents a solution based on scalable coding. presented improves shortages caused by network, consequently, providing better service adopting coding scheme.