A5070969455- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Advanced Image and Video Retrieval Techniques
- Image Processing Techniques and Applications
- Image Enhancement Techniques
- Network Traffic and Congestion Control
- Advanced Optical Network Technologies
- Image Retrieval and Classification Techniques
- Peer-to-Peer Network Technologies
- Robotics and Sensor-Based Localization
- Industrial Vision Systems and Defect Detection
- Domain Adaptation and Few-Shot Learning
- Advanced Photonic Communication Systems
- Advanced Neural Network Applications
- Complex Network Analysis Techniques
- Advanced Image Processing Techniques
- Machine Learning and Data Classification
- Optical Network Technologies
- Software-Defined Networks and 5G
- Optical measurement and interference techniques
- 3D Surveying and Cultural Heritage
- Energy Efficient Wireless Sensor Networks
- COVID-19 diagnosis using AI
- Fire Detection and Safety Systems
- Education Discipline and Inequality
University of Surrey
1986-2023
Signal Processing (United States)
2019-2022
University College London
2001-2007
Nova Southeastern University
2005
This paper summarizes the results of first Monocular Depth Estimation Challenge (MDEC) organized at WACV2023. challenge evaluated progress self-supervised monocular depth estimation on challenging SYNS-Patches dataset. The was CodaLab and received submissions from 4 valid teams. Participants were provided a devkit containing updated reference implementations for 16 State-of-the-Art algorithms novel techniques. threshold acceptance techniques to outperform every one SotA baselines. All...
In the current monocular depth research, dominant approach is to employ unsupervised training on large datasets, driven by warped photometric consistency. Such approaches lack robustness and are unable generalize challenging domains such as nighttime scenes or adverse weather conditions where assumptions about consistency break down. We propose DeFeat-Net (Depth & Feature network), an simultaneously learn a cross-domain dense feature representation, alongside robust depth-estimation...
This paper investigates the challenges for developing current local area network (LAN)-based Ethernet protocol into a technology future architectures that is capable of satisfying dynamic traffic demands with hard service guarantees using high-bit-rate channels (80...100 Gb/s). The objective to combine high-speed optical transmission and physical interfaces (PHY) medium access control (MAC) protocol, designed meet in metropolitan-area networks (MANs). an ideal candidate extension MAN as it...
"Like night and day" is a commonly used expression to imply that two things are completely different. Unfortunately, this tends be the case for current visual feature representations of same scene across varying seasons or times day. The aim paper provide dense representation can perform localization, sparse matching image retrieval, regardless seasonal temporal appearance. Recently, there have been several proposed methodologies deep learning representations. These methods make use ground...
Self-supervised monocular depth estimation (SS-MDE) has the potential to scale vast quantities of data. Unfortunately, existing approaches limit themselves automotive domain, resulting in models incapable generalizing complex environments such as natural or indoor settings.To address this, we propose a large-scale SlowTV dataset curated from YouTube, containing an order magnitude more data than datasets. contains 1.7M images rich diversity environments, worldwide seasonal hiking, scenic...
This paper presents an open and comprehensive framework to systematically evaluate state-of-the-art contributions self-supervised monocular depth estimation. includes pretraining, backbone, architectural design choices loss functions. Many papers in this field claim novelty either architecture or formulation. However, simply updating the backbone of historical systems results relative improvements 25%, allowing them outperform majority existing systems. A systematic evaluation was not...
This paper discusses the results for second edition of Monocular Depth Estimation Challenge (MDEC). was open to methods using any form supervision, including fully-supervised, self-supervised, multi-task or proxy depth. The challenge based around SYNS-Patches dataset, which features a wide diversity environments with high-quality dense ground-truth. includes complex natural environments, e.g. forests fields, are greatly underrepresented in current benchmarks.The received eight unique...
Abstract While there has been considerable attention given to gender differences in the use of computers both at school and home, relatively little published about types teenagers make computers. This paper reports findings usage from a sample 1747 14–18 year olds. The data suggest that while games playing is by far most popular use, it may serve an interest maintenance function facilitates progression on more complete computer facilities learning programming languages. implications these...
How do computers and intelligent agents view the world around them? Feature extraction representation constitutes one basic building blocks towards answering this question. Traditionally, has been done with carefully engineered hand-crafted techniques such as HOG, SIFT or ORB. However, there is no ``one size fits all'' approach that satisfies all requirements. In recent years, rising popularity of deep learning resulted in a myriad end-to-end solutions to many computer vision problems. These...
In this paper we introduce the concepts of Network Planes and Parallel Internets, with objective designing implementing a lightweight solution for viable end-to-end QoS provisioning. The proposed can be deployed very small incremental additions to existing best-effort Internet. Through Plane engineering interconnection, mainly by means intra- inter-domain routing differentiation, service differentiation across Internet achieved.
It has been previously demonstrated [Carlos Da Costa, April 2002] that Internet topologies which were once considered unstructured networks with no global design processes actually follow power-laws, both at the router level and AS (autonomous system) domain level. This discovery very wide implications on network research as well protocol design. The is not only instance to exhibit power laws however; in this paper we present evidence for similar also existing transport layer topologies;...
Self-supervised learning is the key to unlocking generic computer vision systems. By eliminating reliance on ground-truth annotations, it allows scaling much larger data quantities. Unfortunately, self-supervised monocular depth estimation (SS-MDE) has been limited by absence of diverse training data. Existing datasets have focused exclusively urban driving in densely populated cities, resulting models that fail generalize beyond this domain. To address these limitations, paper proposes two...
This paper discusses the results of third edition Monocular Depth Estimation Challenge (MDEC). The challenge focuses on zero-shot generalization to challenging SYNS-Patches dataset, featuring complex scenes in natural and indoor settings. As with previous edition, methods can use any form supervision, i.e. supervised or self-supervised. received a total 19 submissions outperforming baseline test set: 10 among them submitted report describing their approach, highlighting diffused foundational...
Summary form only given. This work proposes that wireless signals can be monitored for potential intruders based on signal-sensing, and presents a framework methodology implementing such system. By identifying threats in this way, actions could taken early before the intruder has had opportunity to compromise network. On leading edge of intrusion detection, focus research, is application intelligent technology predict patterns anomalies may point deviant behavior. The specific goal proposal...
Recent approaches to multi-task learning (MTL) have focused on modelling connections between tasks at the decoder level. This leads a tight coupling tasks, which need retraining if new task is inserted or removed. We argue that MTL stepping stone towards universal feature (UFL), ability learn generic features can be applied without retraining.We propose Medusa realize this goal, designing heads with dual attention mechanisms. The shared masks relevant backbone for each task, allowing it...
This paper discusses the results for second edition of Monocular Depth Estimation Challenge (MDEC). was open to methods using any form supervision, including fully-supervised, self-supervised, multi-task or proxy depth. The challenge based around SYNS-Patches dataset, which features a wide diversity environments with high-quality dense ground-truth. includes complex natural environments, e.g. forests fields, are greatly underrepresented in current benchmarks. received eight unique...
Self-supervised monocular depth estimation (SS-MDE) has the potential to scale vast quantities of data. Unfortunately, existing approaches limit themselves automotive domain, resulting in models incapable generalizing complex environments such as natural or indoor settings. To address this, we propose a large-scale SlowTV dataset curated from YouTube, containing an order magnitude more data than datasets. contains 1.7M images rich diversity environments, worldwide seasonal hiking, scenic...