A5074902814- Image Processing Techniques and Applications
- Anomaly Detection Techniques and Applications
- Advanced Image Processing Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Engineering Diagnostics and Reliability
- Brain Tumor Detection and Classification
- Advanced Measurement and Detection Methods
- Software Reliability and Analysis Research
- EEG and Brain-Computer Interfaces
- Machine Learning and Data Classification
- Heart Rate Variability and Autonomic Control
- Robot Manipulation and Learning
- Data Visualization and Analytics
- Semantic Web and Ontologies
- Emotion and Mood Recognition
- Biomedical Text Mining and Ontologies
- Image and Signal Denoising Methods
- Evolutionary Algorithms and Applications
- Human Pose and Action Recognition
- Reinforcement Learning in Robotics
- Satellite Communication Systems
- Gear and Bearing Dynamics Analysis
- Vehicle Routing Optimization Methods
- Advanced Graph Neural Networks
Northeastern University
2024
Southern University of Science and Technology
2023
Hohai University
2022
Pennsylvania State University
2019
Signed Graph Neural Networks (SGNNs) have been shown to be effective in analyzing complex patterns real-world situations where positive and negative links coexist. However, SGNN models suffer from poor explainability, which limit their adoptions critical scenarios that require understanding the rationale behind predictions. To best of our knowledge, there is currently no research work on explainability models. Our goal address decision-making for downstream task link sign prediction specific...
Medical imaging segmentation plays a significant role in the automatic recognition and analysis of lesions. State-of-the-art methods, particularly those utilizing transformers, have been prominently adopted 3D semantic due to their superior performance scalability generalizability. However, plain vision transformers encounter challenges neglect local features high computational complexity. To address these challenges, we introduce three key contributions: Firstly, proposed SegStitch, an...
This paper presents and experimentally validates a concept of end-to-end imitation learning for autonomous systems by using composite architecture convolutional neural network (ConvNet) Long Short Term Memory (LSTM) network. In particular, spatio-temporal deep is developed, which learns to imitate the policy used human supervisor drive car-like robot in maze environment. The spatial temporal components model are learned recurrent architectures, respectively. as function laser light detection...
Signed Graph Neural Networks (SGNNs) have been shown to be effective in analyzing complex patterns real-world situations where positive and negative links coexist. However, SGNN models suffer from poor explainability, which limit their adoptions critical scenarios that require understanding the rationale behind predictions. To best of our knowledge, there is currently no research work on explainability models. Our goal address decision-making for downstream task link sign prediction specific...
Blind image decomposition aims to decompose all components present in an image, typically used restore a multi-degraded input image. While fully recovering the clean is appealing, some scenarios, users might want retain certain degradations, such as watermarks, for copyright protection. To address this need, we add controllability blind process, allowing enter which types of degradation remove or retain. We design architecture named controllable network. Inserted middle U-Net structure, our...
Deep metric learning is often used to learn an embedding function that captures the semantic differences within a dataset. A key factor in many problem domains how this generalizes new classes of data. In observing triplet selection strategies for Metric Learning, we find best performance consistently arises from approaches focus on few, well selected triplets.We introduce visualization tools illustrate beyond measuring accuracy validation data, and behavior range strategies.
Large variations in the scale of visual entities make it a challenge to reduce computational complexity image-feature extraction. To address this challenge, paper presents an extraction method based on multi-scale attention mechanism(IEMAM). The first introduces hierarchical image shrinkage strategy, which uses linear projection control size features at each layer. Next, multi-head module dimensional trimming is proposed complexity. Finally, IEMAM constructs aggregation structure aggregate...