A5063682950- Neural Networks and Applications
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Medical Image Segmentation Techniques
- Cultural Heritage Materials Analysis
- Human Pose and Action Recognition
- Hand Gesture Recognition Systems
- Brain Tumor Detection and Classification
- Advanced Image and Video Retrieval Techniques
- Face recognition and analysis
- Face and Expression Recognition
- Conservation Techniques and Studies
- Video Surveillance and Tracking Methods
- Video Analysis and Summarization
- Advanced Neural Network Applications
- Image Processing and 3D Reconstruction
- Currency Recognition and Detection
- Anomaly Detection Techniques and Applications
- Multimedia Communication and Technology
- Color Science and Applications
- Music and Audio Processing
- Infrared Target Detection Methodologies
- Digital Media Forensic Detection
- Image and Signal Denoising Methods
- Industrial Vision Systems and Defect Detection
University of Westminster
2013-2025
University of Alicante
2005
Queen Mary University of London
1989-2003
Deep Learning (DL), a groundbreaking branch of Machine (ML), has emerged as driving force in both theoretical and applied Artificial Intelligence (AI). DL algorithms, rooted complex non-linear artificial neural systems, excel at extracting high-level features from data. demonstrated human-level performance real-world tasks, including clinical diagnostics, unlocked solutions to previously intractable problems virtual agent design, robotics, genomics, neuroimaging, computer vision, industrial...
Recovering the shape of any 3D object using multiple 2D views requires establishing correspondence between feature points at different views. However changes in viewpoint introduce self-occlusions, resulting nonlinear variations and inconsistent features Here we a multi-view model utilising view-dependent constraint without explicit reference to structures. For transformation, adopt Kernel PCA based on Support Vector Machines.
The fast evolution of the digital video technology has opened new areas research. most important aspect will be to develop algorithms perform cataloguing, indexing and retrieval. basic step is find a way for abstraction, as this help us more browsing large set data with sufficient content representation. In paper we present an overview current key-frame extraction algorithms. We propose Entropy-Difference, algorithm that performs spatial frame segmentation. evaluation on several clips....
The early and precise identification of a brain tumour is imperative for enhancing patient's life expectancy; this can be facilitated by quick efficient segmentation in medical imaging. Automatic tools computer vision have integrated powerful deep learning architectures to enable accurate boundary delineation. Our study aims demonstrate improved accuracy higher statistical stability, using datasets obtained from diverse imaging acquisition parameters. This paper introduces novel, fully...
Pain assessment is a critical aspect of healthcare, influencing timely interventions and patient well-being. Traditional pain evaluation methods often rely on subjective reports, leading to inaccuracies disparities in treatment, especially for patients who present difficulties communicate due cognitive impairments. Our contributions are three-fold. Firstly, we analyze the correlations data extracted from biomedical sensors. Then, use state-of-the-art computer vision techniques videos...
This work presents the design of a real-time system to model visual objects with use self-organising networks. The architecture addresses multiple computer vision tasks such as image segmentation, optimal parameter estimation and object representation. We first develop framework for building non-rigid shapes using growth mechanism maps, then we define an number nodes without overfitting or underfitting network based on knowledge obtained from information-theoretic considerations. present...
Measuring camera system performance and associating it directly to image quality is very relevant, whether images are aimed for viewing, or as input machine learning automated recognition algorithms. The Modulation Transfer Function (MTF) a well-established measure evaluating this performance. This study proposes novel methodology measuring MTFs from natural scenes, by adapting the standardized Slanted Edge Method (ISO 12233). method involves edge detection techniques, select extract...
The contrast sensitivity function (CSF) characterizes spatial detection in the human visual system and is typically measured from simple, synthetic stimuli. We used frequency decomposition, RMS modulation, a yes/no paradigm an adaptive staircase to measure isolated contextual CSFs (iCSFs cCSFs) natural images. employed Barten's mechanistic model adapted it for modeling purposes by postulating that, signal given band, when presented amongst other broadband signals, can be modeled as if noise....
Recognition of human behaviours requires modeling the underlying spatial and temporal structures their motion patterns. Such are intrinsically probabilistic therefore should be modelled as stochastic processes. In this paper we introduce a framework to recognise based on both learning prior continuous propagation density models behaviour Prior is learned from training sequences using hidden Markov augmented by current visual observation.
An algorithm is described for modelling and recognising temporal structures of visual activities. The method based on (1) learning prior probabilistic knowledge using hidden Markov models, (2) automatic clustering states expectation maximisation (3) observation augmented conditional density distributions to reduce the number samples required propagation therefore improve recognition speed robustness.
Brain tumour diagnosis is a challenging task yet crucial for planning treatments to stop or slow the growth of tumour. In last decade, there has been dramatic increase in use convolutional neural networks (CNN) their high performance automatic segmentation tumours medical images. More recently, Vision Transformer (ViT) become central focus imaging its robustness and efficiency when compared CNNs. this paper, we propose novel 3D transformer named CATBraTS brain semantic on magnetic resonance...
This paper aims to address the ability of self-organizing neural network models manage real-time applications. Specifically, we introduce fAGNG (fast Autonomous Growing Neural Gas), a modified learning algorithm for incremental model Gas (GNG) network. The with its attributes growth, flexibility, rapid adaptation, and excellent quality representation input space makes it suitable real time However, under constraints GNG fails produce optimal topological map any data set. In contrast existing...