A5037667602- Multimodal Machine Learning Applications
- Medical Image Segmentation Techniques
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Robotics and Sensor-Based Localization
- Topic Modeling
- Medical Imaging and Analysis
- Image Processing and 3D Reconstruction
- Augmented Reality Applications
- Generative Adversarial Networks and Image Synthesis
- Image and Signal Denoising Methods
- Image and Object Detection Techniques
- Industrial Vision Systems and Defect Detection
- Transportation and Mobility Innovations
- Acute Ischemic Stroke Management
- Advanced X-ray and CT Imaging
- Advanced Image Processing Techniques
- Soft Robotics and Applications
- Image Enhancement Techniques
- Advanced Image and Video Retrieval Techniques
- Retinal Imaging and Analysis
University of Liverpool
2021-2024
University of New Brunswick
2004
A simple and cost-effective real-time image processing system has been developed to enable rapid prototyping of full-fledged systems. Images are periodically acquired from a video source using specialized control software. Image data is passed via interprocess communication modules executed in the MATLAB environment. Processed results returned software displayed real-time. have for recognition visual enhancement gastrointestinal polyps contained endoscope video. These implemented within...
Endovascular robots have been actively developed in both academia and industry. However, progress toward autonomous catheterization is often hampered by the widespread use of closed-source simulators physical phantoms. Additionally, acquisition large-scale datasets for training machine learning algorithms with endovascular usually infeasible due to expensive medical procedures. In this chapter, we introduce CathSim, first open-source simulator intervention address these limitations. CathSim...
Domain Adaptation for semantic segmentation is of vital significance since it enables effective knowledge transfer from a labeled source domain (i.e., synthetic data) to an unlabeled target real images), where no effort devoted annotating samples. Prior adaptation methods are mainly based on image-to-image translation model minimize differences in image conditions between and domain. However, there guarantee that feature representations different classes the can be well separated, resulting...
Endovascular intervention training is increasingly being conducted in virtual simulators. However, transferring the experience from endovascular simulators to real world remains an open problem. The key challenge environments are usually not realistically simulated, especially simulation images. In this paper, we propose a new method translate images simulator X-ray Previous image-to-image translation methods often focus on visual effects and neglect structure information, which critical for...
The problem of unpaired infrared-to-visible image translation has gained significant attention due to its ability generate visible images with color information from low-detail grayscale infrared inputs. However, current methodologies often depend on conventional style transfer techniques, which constrain the spatial resolution output be equivalent that input image. fixed generation pattern results in blurry generated when translating low-resolution inputs, and utilizing high-resolution...
Developing generalizable models that can effectively learn from limited data and with minimal reliance on human supervision is a significant objective within the machine learning community, particularly in era of deep neural networks. Therefore, to achieve data-efficient learning, researchers typically explore approaches leverage more related or unlabeled without necessitating additional manual labeling efforts, such as Semi-Supervised Learning (SSL), Transfer (TL), Data Augmentation (DA)....
We introduce a shape-sensitive loss function for catheter and guidewire segmentation utilize it in vision transformer network to establish new state-of-the-art result on large-scale X-ray images dataset. transform network-derived predictions their corresponding ground truths into signed distance maps, thereby enabling any networks concentrate the essential boundaries rather than merely overall contours. These SDMs are subjected transformer, efficiently producing high-dimensional feature...
The development of vision models for real-world applications is hindered by the challenge annotated data scarcity, which has necessitated adoption dataefficient visual learning techniques such as semi-supervised learning. Unfortunately, prevalent cross-entropy supervision limited its focus on category discrimination while disregarding semantic connection between concepts, ultimately results in suboptimal exploitation scarce labeled data. To address this issue, paper presents a novel approach...