A5079767092- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Advanced Vision and Imaging
- Cardiac Valve Diseases and Treatments
- Machine Learning in Materials Science
- Advanced Optical Imaging Technologies
- Advanced Image Fusion Techniques
- Metal-Organic Frameworks: Synthesis and Applications
- Cardiovascular Function and Risk Factors
- X-ray Diffraction in Crystallography
- Cancer-related molecular mechanisms research
- Image and Video Quality Assessment
- Infrastructure Maintenance and Monitoring
- Topic Modeling
- Advanced MRI Techniques and Applications
- Natural Language Processing Techniques
- Advanced Text Analysis Techniques
University of Chinese Academy of Sciences
2022-2024
Institute of Information Engineering
2022-2024
Chinese Academy of Sciences
2022-2024
Northeast Forestry University
2023
Document-level Event Extraction aims to identify events from an entire article. It is quite a challenging task because event arguments scatter across several sentences and multiple in document may have influence on each other. Previous methods, however, did not take advantage of structures that been proved be effective for sentence-level extraction. In this work, we propose structure-aware heterogeneous graph with subsentences document-level Firstly, build syntactic capture long-range...
This paper presents a novel framework for converting 2D videos to immersive stereoscopic 3D, addressing the growing demand 3D content in experience. Leveraging foundation models as priors, our approach overcomes limitations of traditional methods and boosts performance ensure high-fidelity generation required by display devices. The proposed system consists two main steps: depth-based video splatting warping extracting occlusion mask, stereo inpainting. We utilize pre-trained stable...
Cardiac diseases have high mortality rates and are a significant threat to human health. Echocardiography is commonly used imaging technique diagnose cardiac because of its portability, non-invasiveness low cost. Precise segmentation basic structures crucial for cardiologists efficiently diseases, but this task challenging due several reasons, such as: (1) image contrast, (2) incomplete cardiac, (3) unclear border between the ventricle atrium in some echocardiographic images. In paper, we...
The real-world image degradation in the super-resolution task is recently considered as a combination of Gaussian blur, down-sampling, and additional white noise. To han-dle this degradation, previous methods estimate blur kernel or model based on randomly selected patch. However, these cannot degradations with high-level noise well they ignore spatial variability even existence Moreover, using denoising networks to preprocess low-resolution images also fails due loss important...
Blind super-resolution task aims to restore low-resolution im"ages with unknown degradations high-resolution counter-parts. Existing methods rely on estimation re-construct images. However, they need human involvement obtain the best results as treat types of known conditions and manually select corresponding trained models. Moreover, cannot fully use estimated generate blurry artifacts ignore that impact images is re-lated contents. In this paper, we propose HIS-NEST which contains an...
Zero shot learning aims to recognize objects whose instances may not be covered by the training data. To generalize knowledge from seen classes novel ones, semantic space is built embed various views into multi-modal embeddings. Existing embeddings neglect relationships between which are essential transfer classes. Moreover, existing zero models ignore complementarity different modalities. tackle these problems, in this work, we resort graph theory explicitly model interdependence and then...