Search

The Covid19 pandemic has significantly changed our ways of living. Government authorities around the world have come up with safety regulations to help reduce the spread of this deadly virus. Covering the mouth and nose using facial masks is identified as an effective step to suppress the transmission of the infected droplets from one human to the other. While the usage of facial masks has been a common practice in several Asian societies, this practice is fairly new to the rest of the world including modern western societies. Hence, it can be noticed that the facial masks are either worn incorrectly (or sometimes not worn) by a significant number of people. Given the fact that the majority of the world population is only getting accustomed to this practice, it would be essential for surveillance systems to monitor if the general population is abiding by the regulatory standards of correctly wearing a facial mask. This paper uses deep learning algorithms to track and classify face masks. The research proposes a mask detection model based on Convolutional Neural Networks to discern between a correct usage of facial masks and its incorrect usages or even lack of it. Different architectures have been tested (even on real-time video streams) to obtain the best accuracy of 98.9% over four classes. These four classes include correctly worn, incorrectly worn on the chin, incorrectly worn on mouth and chin, and not wearing a mask at all. The novelty of this work is in the detection of the type of inaccuracy in wearing the face mask rather than just detecting the presence or absence of the same.

Trees are essential for climate change adaptation or even mitigation to some extent. To leverage their potential, effective forest and urban tree management is required. Automated tree detection, localisation, and species classification are crucial to any forest and urban tree management plan. Over the last decade, many studies aimed at tree species classification using aerial imagery yet due to several environmental challenges results were sub-optimal. This study aims to contribute to this domain by first, generating a labelled tree species dataset using Google Maps static API to supply aerial images and Trees In Camden inventory to supply species information, GPS coordinates (Latitude and Longitude), and tree diameter. Furthermore, this study investigates how state-of-the-art deep Convolutional Neural Network models including VGG19, ResNet50, DenseNet121, and InceptionV3 can handle the species classification problem of the urban trees using aerial images. Experimental results show our best model, InceptionV3 achieves an average accuracy of 73.54 over 6 tree species.

Beavers' habitat is known to alter the terrain, providing biodiversity in the area, and recently their lifestyle is linked to climatic changes by reducing greenhouse gases levels in the region. To analyse the impact of beavers’ habitat on the region, it is, therefore, necessary to estimate the terrain alterations caused by beaver actions. Furthermore, such terrain analysis can also play an important role in domains like wildlife ecology, deforestation, land-cover estimations, and geological mapping. Deep learning models are known to provide better estimates on automatic feature identification and classification of a terrain. However, such models require significant training data. Pre-existing terrain datasets (both real and synthetic) like CityScapes, PASCAL, UAVID, etc, are mostly concentrated on urban areas and include roads, pathways, buildings, etc. Such datasets, therefore, are unsuitable for forest terrain analysis. This paper contributes, by providing a finely labelled novel dataset of forest imagery around beavers’ habitat, captured from a high-resolution camera on an aerial drone. The dataset consists of 100 such images labelled and classified based on 9 different classes. Furthermore, a baseline is established on this dataset using state-of-the-art semantic segmentation models based on performance metrics including Intersection Over Union (IoU), Overall Accuracy (OA), and F1 score.

The next generation of many-core enabled large-scale computing systems relies on thousands of billions of heterogeneous processing cores connected to form a single computing unit. In such large-scale computing environments, resource management is one of the most challenging, and complex issues for efficient resource sharing and utilization, particularly as we move toward Future ManyCore Systems (FMCS). This work proposes a novel resource management scheme for future peta-scale many-core-enabled computing systems, based on hybrid adaptive resource discovery, called ElCore. The proposed architecture contains a set of modules which will dynamically be instantiated on the nodes in the distributed system on demand. Our approach provides flexibility to allocate the required set of resources for various types of processes/applications. It can also be considered as a generic solution (with respect to the general requirements of large scale computing environments) which brings a set of interesting features (such as auto-scaling, multitenancy, multi-dimensional mapping, etc,.) to facilitate its easy adaptation to any distributed technology (such as SOA, Grid and HPC many-core). The achieved evaluation results assured the significant scalability and the high quality resource mapping of the proposed resource discovery and management over highly heterogeneous, hierarchical and dynamic computing environments with respect to several scalability and efficiency aspects while supporting flexible and complex queries with guaranteed discovery results accuracy. The simulation results prove that, using our approach, the mapping between processes and resources can be done with high level of accuracy which potentially leads to a significant enhancement in the overall system performance., 46 pages, 18 figures

Doutoramento em Informática Tecnologias de Computação Distribuída em larga escala tais como Cloud, Grid, Cluster e Supercomputadores HPC estão a evoluir juntamente com a emergência revolucionária de modelos de múltiplos núcleos (por exemplo: GPU, CPUs num único die, Supercomputadores em single die, Supercomputadores em chip, etc) e avanços significativos em redes e soluções de interligação. No futuro, nós de computação com milhares de núcleos podem ser ligados entre si para formar uma única unidade de computação transparente que esconde das aplicações a complexidade e a natureza distribuída desses sistemas com múltiplos núcleos. A fim de beneficiar de forma eficiente de todos os potenciais recursos nesses ambientes de computação em grande escala com múltiplos núcleos ativos, a descoberta de recursos é um elemento crucial para explorar ao máximo as capacidade de todos os recursos heterogéneos distribuídos, através do reconhecimento preciso e localização desses recursos no sistema. A descoberta eficiente e escalável de recursos ´e um desafio para tais sistemas futuros, onde os recursos e as infira-estruturas de computação e comunicação subjacentes são altamente dinâmicas, hierarquizadas e heterogéneas. Nesta tese, investigamos o problema da descoberta de recursos no que diz respeito aos requisitos gerais da escalabilidade arbitrária de ambientes de computação futuros com múltiplos núcleos ativos. A principal contribuição desta tese ´e a proposta de uma entidade de descoberta de recursos adaptativa híbrida (Hybrid Adaptive Resource Discovery - HARD), uma abordagem de descoberta de recursos eficiente e altamente escalável, construída sobre uma sobreposição hierárquica virtual baseada na auto-organizaçãoo e auto-adaptação de recursos de processamento no sistema, onde os recursos computacionais são organizados em hierarquias distribuídas de acordo com uma proposta de modelo de descriçãoo de recursos multi-camadas hierárquicas. Operacionalmente, em cada camada, que consiste numa arquitetura ponto-a-ponto de módulos que, interagindo uns com os outros, fornecem uma visão global da disponibilidade de recursos num ambiente distribuído grande, dinâmico e heterogéneo. O modelo de descoberta de recursos proposto fornece a adaptabilidade e flexibilidade para executar consultas complexas através do apoio a um conjunto de características significativas (tais como multi-dimensional, variedade e consulta agregada) apoiadas por uma correspondência exata e parcial, tanto para o conteúdo de objetos estéticos e dinâmicos. Simulações mostram que o HARD pode ser aplicado a escalas arbitrárias de dinamismo, tanto em termos de complexidade como de escala, posicionando esta proposta como uma arquitetura adequada para sistemas futuros de múltiplos núcleos. Também contribuímos com a proposta de um regime de gestão eficiente dos recursos para sistemas futuros que podem utilizar recursos distribuíos de forma eficiente e de uma forma totalmente descentralizada. Além disso, aproveitando componentes de descoberta (RR-RPs) permite que a nossa plataforma de gestão de recursos encontre e aloque dinamicamente recursos disponíeis que garantam os parâmetros de QoS pedidos. Large scale distributed computing technologies such as Cloud, Grid, Cluster and HPC supercomputers are progressing along with the revolutionary emergence of many-core designs (e.g. GPU, CPUs on single die, supercomputers on chip, etc.) and significant advances in networking and interconnect solutions. In future, computing nodes with thousands of cores may be connected together to form a single transparent computing unit which hides from applications the complexity and distributed nature of these many core systems. In order to efficiently benefit from all the potential resources in such large scale many-core-enabled computing environments, resource discovery is the vital building block to maximally exploit the capabilities of all distributed heterogeneous resources through precisely recognizing and locating those resources in the system. The efficient and scalable resource discovery is challenging for such future systems where the resources and the underlying computation and communication infrastructures are highly-dynamic, highly-hierarchical and highly-heterogeneous. In this thesis, we investigate the problem of resource discovery with respect to the general requirements of arbitrary scale future many-core-enabled computing environments. The main contribution of this thesis is to propose Hybrid Adaptive Resource Discovery (HARD), a novel efficient and highly scalable resource-discovery approach which is built upon a virtual hierarchical overlay based on self-organization and self-adaptation of processing resources in the system, where the computing resources are organized into distributed hierarchies according to a proposed hierarchical multi-layered resource description model. Operationally, at each layer, it consists of a peer-to-peer architecture of modules that, by interacting with each other, provide a global view of the resource availability in a large, dynamic and heterogeneous distributed environment. The proposed resource discovery model provides the adaptability and flexibility to perform complex querying by supporting a set of significant querying features (such as multi-dimensional, range and aggregate querying) while supporting exact and partial matching, both for static and dynamic object contents. The simulation shows that HARD can be applied to arbitrary scales of dynamicity, both in terms of complexity and of scale, positioning this proposal as a proper architecture for future many-core systems. We also contributed to propose a novel resource management scheme for future systems which efficiently can utilize distributed resources in a fully decentralized fashion. Moreover, leveraging discovery components (RR-RPs) enables our resource management platform to dynamically find and allocate available resources that guarantee the QoS parameters on demand.

Doutoramento em Informática Submitted by Manuel Jesus (albertojesus@ua.pt) on 2017-10-30T11:56:48Z No. of bitstreams: 1 Resource discovery for arbitrary scale systems.pdf: 10658800 bytes, checksum: 26d82020a978963c9f80a32122b3ad48 (MD5) Made available in DSpace on 2017-10-30T11:56:48Z (GMT). No. of bitstreams: 1 Resource discovery for arbitrary scale systems.pdf: 10658800 bytes, checksum: 26d82020a978963c9f80a32122b3ad48 (MD5) Previous issue date: 2017