A5018060609- Music and Audio Processing
- Anomaly Detection Techniques and Applications
- Electrical Contact Performance and Analysis
- Advanced Vision and Imaging
- Infrastructure Maintenance and Monitoring
- Advanced Image Processing Techniques
- Traditional Chinese Medicine Studies
- Evacuation and Crowd Dynamics
- BIM and Construction Integration
- Railway Systems and Energy Efficiency
- Image and Signal Denoising Methods
- Video Surveillance and Tracking Methods
- COVID-19 diagnosis using AI
- Advanced Neuroimaging Techniques and Applications
- Blockchain Technology Applications and Security
- Identification and Quantification in Food
- Electromagnetic Compatibility and Noise Suppression
- Image Enhancement Techniques
- Genetics, Bioinformatics, and Biomedical Research
- 3D Surveying and Cultural Heritage
- Fractal and DNA sequence analysis
- Occupational Health and Safety Research
- AI in cancer detection
- Coral and Marine Ecosystems Studies
- Remote-Sensing Image Classification
Southwest Jiaotong University
2023-2025
Guilin University of Electronic Technology
2021-2023
Tongji University
2023
Tsinghua University
2022
South China University of Technology
2022
Chongqing University
2022
Thanks to the development of deep learning, machine abnormal sound detection (MASD) based on unsupervised learning has exhibited excellent performance. However, in task MASD, there are discrepancies between acoustic characteristics test set and training under physical parameter changes (domain shifts) same machine's operating conditions. Existing methods not only struggle stably learn signal features various domain shifts but also inevitably increase computational overhead. To address these...
To alleviate the social contradiction between limited medical resources and increasing needs, image-assisted diagnosis based on deep learning has become research focus in Wise Information Technology of med. Most existing segmentation models Convolution or Transformer have achieved relatively sound effects. However, Convolution-based model with a receptive field cannot establish long-distance dependencies features as Network deepens. The Transformer-based produces large computation overhead...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
ABSTRACT Camouflaged object detection (COD) aims to detect objects that ‘blend in’ with their surroundings and the lack of a clear boundary between target background in COD tasks makes accurate targets difficult. Although many innovative algorithms methods have been developed improve results camouflaged detection, problem poor accuracy complex scenes still exists. To camouflage segmentation, algorithm using contextual feature enhancement an attention mechanism called amplify predict network...
Rainy days usually degrade the visual effect of images and videos. At present, most deraining models for single adopt gradual optimization or elimination to remove rain streaks, but actually with relatively low efficiency in real tasks. An efficient one-stage model, Efficient Transformer Derain Network (ETDNet), is proposed streaks efficiently. A new architecture designed provide rich multiple scales context information, making model extract features a coarse-to-fine way. Multiple expansion...
Intelligent construction (IC) has emerged as a new approach to transforming the architecture, engineering, and (AEC) industry through integration of advanced information technologies such artificial intelligence (AI) Internet Things (IoT). However, due its interdisciplinary nature, relevant documents on IC are diverse fragmented. To provide comprehensive understanding research progress future opportunities in offer suggestions for both developing developed countries, this study employed...
Crowd counting is an important research topic in the fields of computer vision and image processing, with monitoring management crowded scenes becoming increasingly prominent issue.Existing methods still suffer from problem severe overlap density maps within dense areas, leading to inadequate localization accuracy.This paper presents innovative on crowd localization.Firstly, addressing limitations performance existing algorithms, we optimize generation method FIDT maps, decoupling tasks.By...
In the realm of low-resolution (LR) to high-resolution (HR) image reconstruction, denoising diffusion probabilistic models (DDPMs) are recognized for their superior perceptual quality over other generative models, attributed adept handling various degradation factors in LR images, such as noise and blur. However, DDPMs predominantly focus on a single modality super-resolution (SR) reconstruction from thus overlooking rich potential information multimodal data. This lack integration...
In recent years, as the Ether platform has grown by leaps and bounds. Numerous unscrupulous individuals have used illegal transaction to defraud large sums of money, causing billions dollars losses investors worldwide. Facing endless stream based on smart contracts problems, such transaction, money laundering, financial fraud, phishing. Currently, are only detected a single view contract’s contract code feature account feature, which is not incomplete, but also fully representative features....
Mangroves are special vegetation that grows in the intertidal zone of coast and has extremely high ecological environmental value. Different mangrove species exhibit significant differences functions responses, so accurately identifying distinguishing these is crucial for protection monitoring. However, recognition faces challenges, such as morphological similarity, complexity, target size variability, data scarcity. Traditional monitoring methods mainly rely on expensive operationally...
Abstract The coupling of the pantograph-catenary system represents a complex multi-disciplinary challenge, intertwining multi-fields and intricate interactions. Based on thermodynamic theory, this paper establishes thermal-force model for pantograph carbon slider high-speed trains, taking into account multiple physical field factors such as heat, dynamics, electricity. Utilizing contact test bench, we validate full-scale thermo-mechanical slider. Thereafter, assess coupled temperature stress...
As for recognizing Zhuang minority pattern symbols, current recognition models often cause high computational overhead and low accuracy since symbols have large feature vectors some complex features. In this paper, we present the efficient attention receptive field you only look once (Earf-YOLO), a new scheme to address those problems. Firstly, global-local-transformer (GLocalT) structure is proposed, through which other control systems are introduced into axial self-attention module,...
Deep learning advancements have notably improved the performance of unsupervised machine anomalous sound detection. However, a recognized challenge in detection tasks is discrepancies observed signal features test and training sets under identical operating conditions, situation referred to as domain shifts. Existing strategies difficulty consistently across varying shifts, which inevitably escalates computational demands. To address these issues, we put forward Unsuper-TDGCN, an innovative...
Abstract With reference to the limitations of YOLOv3 in recognizing symbols on Zhuang pattern, such as slow detection speed, unable detect small object, and inaccurate positioning bounding box, we propose a new model: Earf-YOLO (Efficient Attention Receptive Field You only look once) this paper. In EarF-YOLO, present an attention module: CBEAM (Convolution Block Efficient Module) at first, which provides feature maps from channel spatial dimensions. module, local cross-channel interaction...