Direção de Curso
Prof. Doutora Manuela Pereira de Sousa (mpereira(at)ubi.pt)
Informação Geral
Segundo Ciclo em Engenharia Informática
Secretariado
Alexandra Isabel Oliveira Ruas (secretariado(at)di.ubi.pt)
+351 275 242081 (ext.: 1601)
+351 275 319899
Morada
Universidade da Beira Interior
Departamento de Informática
Rua Marquês d'Ávila e Bolama
6201-001 Covilhã, Portugal
Regulamento do Grau de Mestre da Universidade da Beira Interior
Regulamento Específico do Mestrado em Engenharia Informática
Decreto-Lei nº63/2016
MS Word: Template_MEI-Thesis-Word.doc
Latex: Template_MEI-dissertacao-LaTex.zip
14453 - Opção 1
14454 - Opção 2
14455 - Opção 3
14456 - Opção 4
11457 - Opção 5
14483 - Opção 6
14484 - Opção 7
14485 - Opção 8
14486 - Opção 9
14487 - Opção 10
14481 - Projeto de Dissertação ou de Estágio em Engenharia Informática
14482 - Tópicos Emergentes em Engenharia Informática
14477 - Dissertação ou Estágio em Engenharia Informática
14450 - Qualidade de Software
14451 - Sistemas de Informação Organizacionais
14452 - Sistemas de Gestão de Bases de Dados
14460 - Linguagens de Programação e Compiladores
14464 - Protocolos de Comunicação
14469 - Aprendizagem Automática
14472 - Computação Interativa e Visualização
14474 - Interfaces Hardware/Software
14459 - Computação na Nuvem
14465 - Segurança de Sistemas Informáticos
14466 - Segurança e Fiabilidade de Software
14467 - Tecnologias de Virtualização e Centros de Dados
14471 - Computação Multimédia
14473 - Inteligência Computacional
14475 - Computação Gráfica em Jogos Digitais
14476 - Visão Computacional
14479 - Ciência de Dados
14480 - Internet das Coisas
| Tipologia. | Título do trabalho. | Resumo do trabalho. | Orientador(a) UBI ou Supervisor(a) Empresa. | Contacto. | + Info. | Local. |
|---|---|---|---|---|---|---|
| Dissertação. | Zero-shot Interpretable Object Recognition | This work comprises the research and development of a novel framework for interpretable object recognition, based in detection/registration deep learning frameworks. The main idea is to start from an object mesh describing the 3D structure of one object and use generative deep learning frameworks to simulate learning instances of the corresponding class. | Hugo Proença | hugomcp@di.ubi.pt | link | UBI |
| Dissertação em contexto empresarial. | EFinHealth - Environment Factors in Health Monitoring | Monitoring the health of a human being can be broader than monitoring patients themselves, as several environmental aspects can also contribute to their health status. Living conditions can pose a significant threat to health when considering factors such as allergens (e.g. mold) that may increase the risk of respiratory diseases, improper ventilation and temperature control that can cause exposure to excessive heat or cold, among others. Access to safe, affordable, and good quality housing is already a challenge that climate change (higher temperatures, excessive precipitation, floods) can aggravate. It is therefore important to monitor, understand and control these factors in order to preserve one's health. This project aims to create a solution that will help people monitor their living environment and access data that they can understand and act upon. The goal of this work is to develop a solution that presents data from their living environment to its users in an understandable and usable way. Therefore, the design of this solution must be guided by user research activities. Different approaches should be explored and the end result should guarantee that users have access to knowledge that they can really use and that can help them act in improving their living conditions. The developed work will contribute to the understanding on how to present data to users in a meaningful and empowering manner and explore a broader definition of health that includes new health indicators such as environmental aspects and living conditions. | Ana Vasconcelos | ana.vasconcelos@aicos.fraunhofer.pt | Fraunhofer Portugal AICOS | |
| Dissertação. | Data agnostic RaaS | Recommendation systems are commonly used in different various decision-making processes (What music? What movie? What book? What product?...). Developers can resource to a number of algorithms, API's and frameworks as a way to reduce the time-to-market when implementing a custom recommendation system. Nevertheless, after implementation, a recommender system has to evolve, to be integrated and to be maintained during its life cycle. This is too costly and relatively few corporations manage these tasks by creating and sustaining a speciffic business unit inside their organizations. On the other hand, it is possible to find an increasing number of go-to-market solutions for Recommender-system as-a-Service (RaaS) allowing the seamless integration (with only a few clicks or lines of code) of a recommender system - for instance into an e-commerce website or into a blog. A RaaS typically is highly scalable and delivers reasonable results after a few hours or even minutes in production - compared to months of development for a custom-made recommender system. However, one of the downsides of a RaaS is that it usually comes with some compromises relating to flexibility, control and effectiveness when compared to a custom-made recommender system. Even though we are seeing more and more of these services, due to their lack of flexibility, they usually lean into a single use case, focusing on sales or video or advertising or any other speciffic use case for recommendations. | Paulo Fazendeiro | pandre@di.ubi.pt | UBI | |
| Estágio. | Otimização e implementação de dois servidores recorrendo ao sistema TrueNAS SCALE | Revisão bibliográfica e estudo dos servidores Supermicro (hardware, BIOS, etc) e sistemas TrueNAS, OpenZFS/RAIDZ/dRAID. Otimização de esquemas de configurações gerais para cada servidor (compreendendo a conjugação de configurações dos vários "módulos" de software), detetando o melhor desempenho (incluindo o desempenho em rede, transferências de dados, segurança de dados, etc). Colocação em produção em cada servidor do respetivo esquema que regista o melhor desempenho. Preparação de documentação técnica e científica. | Emanuel Maldonado | e.maldonado@fcsaude.ubi.pt | link | CICS-Centro de Investigação em Ciências da Saúde, Universidade da Beira Interior |
| Dissertação. | Identificação de perdas técnicas em redes de energia elétrica | O trabalho deverá incidir sobre o controlo e a redução de perdas técnicas (energia elétrica perdida ao longo da cadeia de produção, transporte, transformação e distribuição) nas redes de distribuição de energia elétrica de baixa tensão, alimentadas por centrais termoelétricas e por centrais de energia proveniente de fontes renováveis. Através da identificação de padrões de consumo e outras informações, provenientes das bases de dados das concessionárias de energia e das agências reguladoras do sector elétrico, pretende-se identificar situações críticas, em tempo real, suscetíveis de provocar perdas técnicas nas redes de distribuição de energia elétrica. O trabalh passa por comparar técnicas computacionais, baseadas em Aprendizagem Automática, com o objetivo de estabelecer uma metodologia que permita classificar e selecionar características dos padrões de comportamento das redes de energia elétrica. | Paulo Fazendeiro | pandre@di.ubi.pt | UBI | |
| Dissertação. | Técnicas de aprendizagem automática para tipificação do risco de abandono escolar | O objetivo geral passa por criar um modelo de aprendizagem automática (machine learning) capaz de prever problemas como (i) a evasão dos alunos inscritos e (ii) o aumento da taxa de reprovações em instituições do ensino superior. Pretende-se utilizar técnicas de clustering para analisar dados educacionais e identificar padrões e tendências na relação entre o desempenho académico, a vida socioeconómica e a qualidade da instituição de ensino. Os resultados obtidos podem fornecer insights adicionais sobre como esses fatores afetam o desempenho académico e contribuir para a compreensão das relações entre esses fatores. | Paulo Fazendeiro | pandre@di.ubi.pt | UBI | |
| Dissertação. | Identificação e seguimento de espécies ameaçadas em imagens aéreas | Pretende-se desenvolver e implementar um sistema capaz de assistir instituições de conservação da natureza na localização de espécies ameaçadas com maior rapidez. O objetivo é tornar a tarefa de proteção de animais marinhos mais rápida e eficaz. Protegendo contra predadores, de caçadores furtivos ou de outras ameaças. Neste caso concreto pretende-se criar um sistema capaz de detectar tartarugas marinhas em imagens recolhidas por drone. A principal ameaça que este tipo de animais enfrenta, são a captura acidental na pesca, mas existem muitas outras ameaças. Por exemplo a poluição por plásticos marinhos, o aumento do nível da água do mar induzido pelo clima, a erosão das praias, a caça ilegal dos seus ovos, a iluminação artificial pode interferir na orientação das tartarugas, os detritos marinhos podem prender as tartarugas tal como os derrames de óleo. O sistema a desenvolver tem potencial para ajudar a lidar com várias destas situações. | Paulo Fazendeiro | pandre@di.ubi.pt | UBI | |
| Dissertação. | Word Embeddings by Generality | The focus of this master's thesis is to propose developing a novel approach to asymmetric word embeddings utilizing AIS. The aim is to explore how AIS can be effectively leveraged to enhance the representation of words in a manner that captures their asymmetric relationships. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Dissertação. | Black Box to Mining of Massive Datasets | The main objective of this master's thesis is to propose the development of an abstract "black box" package in Python that integrates Spark machine learning capabilities. This package aims to provide a simplified and user-friendly interface for utilizing Spark's powerful machine-learning algorithms and tools. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Dissertação. | Emotion Retrieval in Mental Health Context | This work proposes an innovative approach to categorizing emotions within the mental health context, aiming to advance our understanding of emotional experiences and their implications for mental well-being. The research significantly contributes to emotion analysis and its application in mental health contexts by leveraging cutting-edge methodologies. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Dissertação. | Personality Retrieval in Mental Health Context | Personality detection and recognition from text is an emerging research field closely tied to sentiment analysis and emotion detection. While sentiment analysis focuses on identifying the positive, neutral, or negative sentiments expressed in text, and emotion analysis aims to detect and recognize specific types of feelings; personality detection deals with identifying an individual's personality traits based on their written expressions. The Big Five personality traits framework is commonly used for formal description in personality detection. The Big Five personality traits are five broad dimensions that capture various aspects of human personality. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Dissertação. | Bert Embeddings | BERT (Bidirectional Encoder Representations from Transformers) embeddings are a powerful technique in natural language processing (NLP) for representing words, phrases, and sentences. BERT embeddings are based on transformer architecture, a type of neural network that excels at capturing contextual information. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Dissertação. | A Web 3D Avatar for Speech | Neste projeto pretende-se desenvolver um avatar 3D para a Web que permita falar com o utilizador, ou seja, através do texto/som que lhe é fornecido, por exemplo através de WebRTC. Assim, é necessário criar um modelo 3D com um conjunto de animações faciais, as quais têm de ser sincronizadas depois com o som que lhe for fornecido. O projeto é para ser desenvolvido em Unity. | Frutuoso Silva | fsilva@di.ubi.pt | link | Regain - UBI |
| Estágio. | Design of a Requirement Analyser and Test Cases Generator | O âmbito deste projecto é o desenvolvimento de uma ferramenta que automatize algumas tarefas de verificação de sistemas para sinalização ferroviária desenvolvidos pela EFACEC. As tarefas são a análise de requisitos, a definição dos casos de teste, a implementação dos cenários de teste e finalmente a execução dos cenários de teste. Existem dois grandes objectivos para este trabalho: (1) Implementar o módulo que analise os requisitos escritos em linguagem natural e verifique a sua compleção e coerência. (2) Desenhar e implementar o módulo que analise os requisitos e define os casos de teste que verifiquem esses requisitos. A ferramenta deverá também criar os cenários de teste "traduzindo" os casos de teste em linguagem capaz de ser executada pela ferramenta onde foi desenvolvido o SW alvo dos testes. | Cláudio Sagres | claudio.sagres@efacec.com | link | EFACEC (possibilidade de trabalho remoto) |
| Dissertação. | ForestVision - Forest Species Health Estimation and Growth | The ForestVision project encompasses the development of an online platform for monitoring, measuring, analyzing and intelligent evaluation of forests to allow early warning systems as well as sustainable planning and management. The project will be based on the use of artificial intelligence and computer vision and focused on the forests of the Iberic Peninsula, starting with the Beiras and Serra da Estrela area and the Beira Baixa in Portugal, as well as the Estremadura region in Spain. The platform will use data of different sources like satellite and drone imagery of different types, like RGB, LIDAR and spectral imagery. The student will have the support of external partners such as environmental agencies to aid in the understanding of the use cases, as well as in the construction of datasets. The goal of this thesis is to develop a computer vision solution to allow the tracking of multiple conditions and scenarios of different flora in a forest ecosystem and adapt the existing platform and computer vision models to real-world use cases. In preparation, the student will review the state-of-the-art forest management systems and sensors used in these use cases, as well as the latest innovations in computer vision techniques, with a specific focus on the following use case needs: 1. Tree health estimation using aerial data; 2. Forest growth, deforestation, and invasive species propagation using segmentation techniques; 3. Tree canopy cover/growth and light penetration estimation using LIDAR or alternative sources. These various types of implementations will be studied and a comparative study will be conducted, to build a guide for how technologies and sensors can be deployed in this specific use case. This work will then be implemented in a platform with a defined front-end and back-end comprising a just-in-time ML model deployment system, so it is expected to be thoroughly documented. The thesis proposal includes a student grant, which adheres to the grant value outlined in accordance with the FCT table. | Luís Alexandre e António Abreu (empresa Zirak) | luis.alexandre@ubi.pt | link | UBI |
| Dissertação. | Análise da Capacidade e Treino de Bots de Conversação na Geração de Regras de Firewall ou de Sistemas de Deteção de Intrusões | Neste trabalho pretende-se explorar a capacidade dos chat bots baseados em grandes modelos de dados atualmente existentes em traduzir pedidos escritos em linguagem natural para regras de sistemas de firewalls ou sistemas de deteção de intrusões. A ideia é medir o quão bem vários sistemas existentes fazem essa tarefa para diferentes sistemas (sobretudo código aberto) e com diferentes graus de detalhe do pedido (e.g., um pedido sem detalhe e só com o nome do ataque, até ao pedido que refere explicitamente que artefactos endereçar na regra). É também objetivo deste trabalho tentar aperfeiçoes um dos modelos existentes para a tarefa específico antes referida, voltando a testá-lo após esse fine-tunning. | Pedro R. M. Inácio | prmi@ubi.pt | link | Universidade da Beira Interior |
| Dissertação. | Design and Assessment of a Model for Cloud Related Interoperability Open Issues | The main challenge of this proposal is the design of a model/standard or system/prototype that complements existing state-of-the-art solutions for cloud interoperability. These should be mainly focused on the vendor lock-in problem that exists in single or interconnected cloud environments. | Tiago Miguel Carrola Simões | tsimoes@di.ubi.pt | link | Universidade da Beira Interior |
| Dissertação. | Performance Evaluation of Microservices in Serverless Architectures | The main challenge of this is to assess or evaluate the current performance of inter-service communication among microservices deployed under serverless architectures. Specifically, to aim for the identification of open issues and thus the development of a new model or prototype capable of surpassing those open issues. | Tiago Miguel Carrola Simões | tsimoes@di.ubi.pt | link | Universidade da Beira Interior |
| Dissertação. | Black Box to Mining of Massive Text | The main objective of this master's thesis is to propose the development of an abstract "black box" package in Python that integrates text mining algorithms and machine learning capabilities, language-independent. This package aims to provide a simplified and user-friendly interface for utilizing text mining and machine-learning algorithms and tools. This master's thesis focuses on developing an innovative "black box" package in Python that aims to revolutionize the field of text mining and machine learning. The primary goal of this package is to provide a comprehensive and language-independent solution for users to leverage the power of text mining algorithms and machine learning techniques through a simplified and user-friendly interface. | Sebastião Pais | sebastiao@di.ubi.pt | link | Di, UBI |
| Dissertação. | Dynamic Multimodal Fusion Learning in Mental Health Context | This master thesis proposes a novel approach to mental health assessment and intervention by combining and analyzing data from various modalities, including text, speech, physiological signals, and images. By dynamically and adaptively leveraging the complementary nature of these modalities, researchers can obtain a more comprehensive understanding of an individual's mental health state, leading to the development of personalized interventions. | Sebastião Pais | sebastiao@di.ubi.pt | link | DI, UBI |
| Estágio. | Estágio na área system engineer | . instalação e configuração de servidores (Windows preferencial) . configuração de redes (servidores) . administração básica de sistemas (Windows preferencial) . aplicação de patches de segurança . configuração de acessos . análise de logs | José Rocha | jose.rocha@itcenter.com | ITCenter (Shar) | |
| Dissertação. | Super-resolution Satellite Imagery for Crop Health Monitoring | The world population growth and the concerns on the sustainability have raised the interest on precision agriculture. In the last years, several solutions have been proposed for image-based crop monitoring using drones, allowing the identification of crop health condition, lack of moisture, weed overgrowth, and crop diseases. Nevertheless, these devices are not always autonomous and restricted to medium size areas. The need to extend this analysis to larger scales without requiring expensive hardware has fostered the interest on exploiting satellite data for this purpose [1,2]. However, when compared with images acquired by drones, these data has a significant lower resolution, decreasing its applicability. This dissertation intends to address this problem by developing a super-resolution method capable of producing images with a reasonable resolution for monitoring soil and plant health from agricultural crops using remote sensing imagery from the European Space Agency. For this, a set of images acquired from Unmanned Aerial Vehicles (UAVs) will be matched with the corresponding satellite images, such that these data is used to train a method capable of approximating the resolution of satellite images to UAVs. | João Neves | jcneves@di.ubi.pt | link | Universidade da Beira Interior |
| Dissertação. | Cross-Dataset Deep Fake Detection | The emergence of deep fake data has raised new challenges to digital media authenticity, creating the need for robust detection methods. While the state-of-the-art approaches can perfectly distinguish between veridical and fake samples obtained from the same dataset where these methods were trained, they fail to attain similar performance in the setting of cross-dataset deep fake detection. In this scenario, the challenge lies in developing techniques that can generalize well across different datasets with varying characteristics and manipulation techniques. This dissertation proposal aims to investigate the problem of cross-dataset deep fake detection, focusing on the development of an innovative detection strategy. In particular, in this work, the student will explore diverse datasets, analyze state-of-the-art cross-dataset detection methods, and address the challenges of improving detection accuracy and robustness across different image manipulations. | João Neves | jcneves@di.ubi.pt | link | Universidade da Beira Interior |
| Dissertação. | Causal Face: Learning Causal Graphs from Facial Attributes for Explaining Face Recognition Models | Face recognition models have been used in many applications, including security systems, social media platforms, and biometric authentication. However, the black-box nature of these models often raises concerns about their interpretability and potential biases. This proposal focuses on the concept of causal graphs to determine the impact that a set of latent variables (e.g., facial attributes) have on the decisions performed by face recognition models. The study aims to explore causal relationships between facial attributes and recognition outcomes, providing insights about the model reasoning and identifying factors that influence model predictions. The research will rely on causal inference and graph learning for the development of an algorithm capable of determining the causal graph representing the decision process of the model based on a set of known facial attributes. | João Neves | jcneves@di.ubi.pt | link | Universidade da Beira Interior |
| Estágio. | ML Orchestrator: Development and Optimization of Machine Learning Pipelines and Platforms | This work aims at developing and optimizing Machine Learning pipelines and platforms. The student will utilize the company's existing local and cloud infrastructure, models, datasets, and communication resources to enhance and streamline the Machine Learning pipelines. The goal of this work is to develop a comprehensive local and cloud solution that is optimized to effectively cater to the specific needs of users and clients., w.r.t. the number of cameras, models and inputs to be analysed and included. The platforms should be constructed based on the existing solution, which already provides the necessary infrastructure to support the development of such architectures. Ultimately, the objective is to create a fully optimized Machine Learning pipeline capable of managing multiple clients and their respective Artificial Intelligence models while effectively orchestrating computational resources. | João Neves | jcneves@di.ubi.pt | link | DeepNeuronic |
Contacto | Alexandra Isabel Oliveira Ruas (secretariado)
Universidade da Beira Interior
Departamento de Informática
Rua Marquês d'Ávila e Bolama
6201-001 Covilhã, Portugal
Coordenadas: N
40°16'38.3", W 7°30'34.6"
Email: secretariado@di.ubi.pt
Phone: +351 275 242081 (ext.: 1601)
Fax: +351 275 319 899