A5065972237- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Advanced Vision and Imaging
- Image Enhancement Techniques
- Blind Source Separation Techniques
- Computer Graphics and Visualization Techniques
- Machine Learning in Materials Science
- Data Quality and Management
- Adversarial Robustness in Machine Learning
- Topic Modeling
- Privacy-Preserving Technologies in Data
- 3D Shape Modeling and Analysis
- Advanced Neural Network Applications
- Image Processing Techniques and Applications
- Geographic Information Systems Studies
- 3D Surveying and Cultural Heritage
- Advanced Chemical Sensor Technologies
- Gaze Tracking and Assistive Technology
- Anomaly Detection Techniques and Applications
- EEG and Brain-Computer Interfaces
- Stochastic Gradient Optimization Techniques
- Cryptography and Data Security
- Computational Drug Discovery Methods
- Natural Language Processing Techniques
Sun Yat-sen University
2023-2024
Southwestern University of Finance and Economics
2023
Shanghai Jiao Tong University
2021
Although recent methods in Unsupervised Domain Adaptation (UDA) have achieved success segmenting rainy or snowy scenes by improving consistency, they face limitations when dealing with more challenging scenarios like foggy and night scenes. We argue that these prior excessively focus on weather-specific features adverse scenes, which exacerbates the existing domain gaps. To address this issue, we propose a new metric to evaluate severity of all offer novel perspective enables task...
Unsupervised Domain Adaptation (UDA) aims to adapt models from labeled source domains unlabeled target domains. When adapting adverse scenes, existing UDA methods fail perform well due the lack of instructions, leading their overlook discrepancies within all scenes. To tackle this, we propose CoDA which instructs distinguish, focus, and learn these at scene image levels. Specifically, consists a Chain-of-Domain (CoD) strategy Severity-Aware Visual Prompt Tuning (SAVPT) mechanism. CoD focuses...
Unsupervised Domain Adaptation (UDA) for semantic segmentation has received widespread attention its ability to transfer knowledge from the source target domains without a high demand annotations. However, under adverse conditions still poses significant challenges autonomous driving, as bad weather observation data may introduce unforeseeable problems. Although previous UDA works are devoted scene tasks, their adaptation process is redundant. For instance, unlabeled snow training must model...
Grounded Multimodal Named Entity Recognition (GMNER) is an emerging information extraction (IE) task, aiming to simultaneously extract entity spans, types, and entity-matched bounding box groundings in images from given sentence-image pairs data. Recent unified methods employing machine reading comprehension (MRC-based) frameworks or sequence generation-based models face challenges understanding the relationships of multimodal entities. MRC-based frameworks, utilizing human-designed queries,...
The field of Remote Sensing Domain Generalization (RSDG) has emerged as a critical and valuable research frontier, focusing on developing models that generalize effectively across diverse scenarios. Despite the substantial domain gaps in RS images are characterized by variabilities such location, wavelength, sensor type, this area remains underexplored: (1) Current cross-domain methods primarily focus Adaptation (DA), which adapts to predefined domains rather than unseen ones; (2) Few...
Data are distributed across different sites due to computing facility limitations or data privacy considerations. Conventional centralized methods—those in which all datasets stored and processed a central facility—are not applicable practice. Therefore, it has become necessary develop learning approaches that have good inference predictive accuracy while remaining free of individual obeying policies regulations protect privacy. In this article, we introduce the basic idea conduct selected...
The low Signal Noise Ratio (SNR) and nonstationarity of electroencephalogram (EEG) signals affect the classification accuracy badly in motor imagery (MI-EEG) classification. In this paper, a Distribution Based Learning (DBL) framework based on deep learning is proposed to improve accuracy. Firstly, uses modified multi band Common Spatial Pattern (CSP) algorithm pre-process raw EEG signals. Secondly, Network (DBLN) utilized divide dataset into two parts. After that, two-step distribution...
Machine learning is becoming a preferred method for the virtual screening of organic materials due to its cost-effectiveness over traditional computationally demanding techniques. However, scarcity labeled data poses significant challenge training advanced machine models. This study showcases potential utilizing databases drug-like small molecules and chemical reactions pretrain BERT model, enhancing performance in materials. By fine-tuning models with from five tasks, version pretrained...
Point cloud classification and segmentation are the primary tasks in 3D computer vision with great application value. Recently, several methods adopt deep neural networks to solve problems by directly taking point clouds as input due their simplicity effectiveness. However, existing only focus on extracting local information of ignores global features which carry importance geometric space. Inspired way humans observe a object (analyzing its overall characteristics first then combining it...