A5059534928- Image Processing Techniques and Applications
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Transportation Planning and Optimization
- Complex Network Analysis Techniques
- Medical Image Segmentation Techniques
- Advanced Neural Network Applications
- Sparse and Compressive Sensing Techniques
- Computational Physics and Python Applications
- Data Management and Algorithms
- Dark Matter and Cosmic Phenomena
- Brain Tumor Detection and Classification
- Data Visualization and Analytics
- Human Mobility and Location-Based Analysis
University of Alberta
2022-2023
York University
2019-2020
Alzheimer's disease is the most prevalent neurodegenerative disorder characterized by degeneration of brain. It classified as a brain causing dementia that presents with memory loss and cognitive impairment. Experts primarily use imaging other tests to rule out disease. To automatically detect patients from healthy controls, this study adopts vision transformer architecture, which can effectively capture global or long-range relationship image features. further enhance network's performance,...
Conventional deconvolution methods utilize hand-crafted image priors to constrain the optimization. While deep-learning-based have simplified optimization by end-to-end training, they fail generalize well blurs unseen in training dataset. Thus, image-specific models is important for higher generalization. Deep prior (DIP) provides an approach optimize weights of a randomly initialized network with single degraded maximum posteriori (MAP), which shows that architecture can serve as prior....
Many deep learning based methods are designed to remove non-uniform (spatially variant) motion blur caused by object and camera shake without knowing the kernel. Some directly output latent sharp image in one stage, while others utilize a multi-stage strategy (e.g. multi-scale, multi-patch, or multi-temporal) gradually restore image. However, these have following two main issues: 1) The computational cost of is high; 2) same convolution kernel applied different regions, which not an ideal...
A comprehensive data analysis system is implemented for the extraction of information and comparison North American public transport systems. The based on network representations systems makes use a span metrics algorithms from established properties in graph theory to complicated domain specific measurements. Due nature big requirement scalability, many heuristic optimizations approximations have been considered system. Integration with other sources specially population density maps also...
Conventional deconvolution methods utilize hand-crafted image priors to constrain the optimization. While deep-learning-based have simplified optimization by end-to-end training, they fail generalize well blurs unseen in training dataset. Thus, image-specific models is important for higher generalization. Deep prior (DIP) provides an approach optimize weights of a randomly initialized network with single degraded maximum posteriori (MAP), which shows that architecture can serve as prior....
Many deep learning based methods are designed to remove non-uniform (spatially variant) motion blur caused by object and camera shake without knowing the kernel. Some directly output latent sharp image in one stage, while others utilize a multi-stage strategy (\eg multi-scale, multi-patch, or multi-temporal) gradually restore image. However, these have following two main issues: 1) The computational cost of is high; 2) same convolution kernel applied different regions, which not an ideal...