3D Vision

3D vision is a promising new tool in manufacturing, diagnosis and entertainment industries. The idea of capturing and presenting content to viewers in 3D is not new. Stereoscopic cameras have existed since the beginning of the 20th century. In the 20’s the idea of stereoscopic TV was started. This native approach at capturing stereoscopic content follows the principles of the human visual system, modeling it with two cameras or multi cameras. Or depth cameras are used to give two-dimensional image source and a depth map. Very recently, image-based 3D reconstruction using convolutional neural networks (CNN) has attracted increasing interest and demonstrated an impressive performance. Given this new era of rapid evolution to estimate the 3D shape of generic objects either from a single or multiple RGB images.

  • Related Bachelor Projects:
    • Matching 2D Sketches to 3D models Using Deep Learning
      Supervised by:
    • Play Bricks: Volume Decomposing in Basic Bricks
      Supervised by:
    • Play Bricks: Creating 3D Facades Model with Special Architectural Component
      Supervised by:
    • 3D Model Construction from Multiple Images (Auto-3D)
      Supervised by:

Augmented, Virtual, and Mixed Reality

Augmented, virtual, and mixed reality applications all aim to enhance a user’s current experience or reality. In this cluster, we will investigate different applications and show how and where these applications could be of benefit.

  • Related Bachelor Projects:
    • Virtually Me
      Supervised by:
    • MINIZORD - VR
      Supervised by:
    • VR for the Visually challenged
      Supervised by:
    • DatAR
      Supervised by:
    • AR4VR - Architecture
      Supervised by:

Autonomous Cars

Autonomous vehicles have held the attention of futurists and technology enthusiasts for some time as evidenced by the continuous research and development in autonomous vehicle technologies over the past two decades. Rapid advances in robotics, artificial intelligence, computer vision, and edge computing capabilities are resulting in machines that can potentially think, see, hear, and move more deftly than humans. Autonomous vehicles in the form of self-driving cars have become the subject of both hype and intense competition among auto majors and technology companies. Self-driving car prototypes that are decked in lidars, radars, cameras, ultrasonic sensors?—?along with heavy computational capabilities under the hood to recognize and maneuver around obstacles?—?is becoming a common sight in many cities. With the emergence of sophisticated autonomous vehicle technologies, we are now on the cusp of their rapid deployment in industrial applications, and the confluence of the Internet of Things (IoT) and AV technologies are poised to re-make and re-imagine industries. Here at the GUC, we started a project to develop a self driving car platform that can drive autonomously around the GUC campus, the topics within this cluster will focus on developing and improving the current developed systems and creating new systems to achieve this goal.

  • Related Bachelor Projects:
    • Visualization Techniques for the In-Vehicle Dashboard
      Supervised by:
    • Detecting road irregularities, speed bumps and obstacles using L
      Supervised by:
    • Detecting road irregularities, speed bumps and obstacles using D
      Supervised by:
    • Drivable Area Detection and Pavement detection using LiDAR, Came
      Supervised by:
    • Intention prediction for pedestrian crossing
      Supervised by:
    • Intention prediction for vehicles
      Supervised by:
    • Software safety for AVs
      Supervised by:
    • Road Mapping for AVs
      Supervised by:
    • Path Planning
      Supervised by:
    • Self-Driving in Egyptian Streets? An Investigation using Simulat
      Supervised by:

Bio-Signal and Image Informatics

Medical Signal and Image Processing enables quantitative analysis and visualization of medical images of numerous modalities such as ECG, EMG, EEG, PET, MRI, or CT of human anatomy.

  • Related Bachelor Projects:
    • Deep Learning Model for 3D Constrctin of Breast Cancer in CT Images
      Supervised by:
    • Deep Learning Utilization in Ultrasound Beamforming Application
      Supervised by:
    • Prediction of seizure likelihood for patients with drug-resistant epilepsy
      Supervised by:
    • X-Ray Multi-class Segmentation Using Deep Learning
      Supervised by:

Brain-Computer Interface Applications

Brain-computer interface (BCI) systems allow controlling computers and other electronic devices using brainwaves, which is relevant to be applied for users especially with reduced motor abilities. BCIs currently provide new hope for people with disabilities whose brain function is still intact. By recording brain activity, BCIs could translate recorded brain activity to actions that can interact with the external environment.

  • Related Bachelor Projects:
    • A Brain-controlled Robotic Arm
      Supervised by:
    • Music Emotion Recognition using Brain Activity
      Supervised by:
    • EEG-based Music Recommendation System
      Supervised by:

Brain-Computer Interfaces for Automotive Applications

It is estimated that traffic accidents represent the fifth leading cause of death in the world, reaching 1.35 million deaths per year. Among the principal causes of the high car-related accidents and mortality are human errors which are largely correlated to distractions, tiredness, or the simultaneous realization of other activities during driving. In Europe, 20% of traffic accidents occur due to diminished levels of vigilance of drivers. Other studies reported that 20% of drivers in Canada (approximately 4.1 million drivers) fall asleep for short intervals at least once while driving in the year preceding the survey. Thus, techniques that could detect when drivers become unfocused could help prevent accidents. Brain-Computer Interfaces (BCIs) could play a leading role in this paradigm given their ability to recognize mental states. Another avenue in which BCIs could be utilized within the automotive domain is using BCIs to control autonomous cars. Given that autonomous cars require minimum intervention from passengers, BCIs could provide means for providing the car with necessary information to complete its trip.

  • Related Bachelor Projects:
    • A Brain-operated Advanced Driver-Assistance System for Emergency Braking
      Supervised by:
    • Sleep Detection using Brain Activity for Driver-Assistance Systems
      Supervised by:
    • A Brain-operated Autonomous Car Controller
      Supervised by:

Character Computing

The rising trend of Adaptive and Seamless Interactive Systems requires having elaborate user profiles to be able to best adapt the application to the user’s behaviour. Instead of focusing on the current state of the user (affect) our interest is to have profiles describing the user’s general state (character). Various models for extracting and adapting to human personalities and character traits exist. This cluster focuses on tackling these issues within the framework of Character Computing. Character Computing is defined as any computing that incorporates the human character within its context. The character is the individual person with all his/her defining or describing features. This includes stable personality traits, variable affective, cognitive and motivational states as well as history, morals, beliefs, skills, appearance and socio-cultural embeddings, to name a few. Character Computing is introduced in a book which will be published by Springer in early 2020. In this book the different topics and problems to be tackled within Character Computing are introduced. One such problem is how data can be collected and managed (--> 2 topics). Data can say a lot about an individual. In the right hands, datasets about diverse aspects of individuals’ lives can go a long way. Knowing enough information about an individual can help draw conclusions to help improve their quality of life, make their interactions with technology better or even improve their wellbeing and prevent them from suffering mental disorders. The first challenge is collecting such datasets containing the required information. The resulting data is very complex and thus requires advanced management and analysis techniques. However, the data is very critical thus privacy and security need to be ensured and maintained at all times. Another problem is how character can be modelled formally in a language similar first order logic( ----> 3 topics). Individuals behave differently within each situation based on their character. This is a complex relationship that needs to be modeled extensively to be able to analyze and understand it. This can only be realized if models from computer science and psychology are combined. Due to the complexity of the task, we are developing an ontology-based model that mainly focuses on the different characteristics of humans and their interaction and effect on performance. An ontology formally represents knowledge as a hierarchy between sets of objects and the relationships between them. The ontology is developed using the Java-based, free, open-source ontology editor and a knowledge management system Protege and is mainly written in the Web Ontology Language (OWL). Description Logic (DL) is used to describe and reason about the concepts of the ontology. The used DL is more expressive than propositional logic and less expressive than first-order logic. Related Reading: - El Bolock, Alia, Abdelrahman, Yomna and Abdennadher, Slim (Eds.). Character Computing. Springer Nature (2020).https://www.springer.com/gp/book/9783030159535 The projects in this cluster are part of a collaboration with the Department of Applied Emotion and Motivation Psychology, Ulm University consisting of multiple workshops and opportunities taking place both at the GUC and Ulm University, that students working on these projects can attend. Very strong technical and programming capabilities are needed for these projects.

  • Related Bachelor Projects:
    • CC-Rules: A Rule Generation and Visualization Platform for the Character Computing Ontology
      Supervised by:
    • CC-DB: Seamless Integration of Multiple Data Sources into the Character Computing Ontology
      Supervised by:
    • Humans: Using Word Semantics and Relations to Automatically Cate
      Supervised by:
    • AppGen 2.0: Extending the AppGen with Data and Knowledge Management
      Supervised by:
    • SAD: Harnessing Self-description and Self-report Data for Detecting Anxiety and Depression
      Supervised by:

Combinatorial Algorithms

According to Kreher and Stinson (1999), combinatorial algorithms are classified into three classes: generation, enumeration, and search algorithms. Generation algorithms "[construct] all the combinatorial structures of a partiular type" (Kreher and Stinson, 1999, p. 1). These include algorithms for generating all subsets, partitions, or all permutations of a set. Enumeration algorithms "[compute the number of different structures of a particular type]" (Kreher and Stinson, 1999, p. 1). For example, finding the number of subsets of size k of a set of size n is an enumeration problem. Search algorithms "[find] at least one example of a structure of a particular type (if it exists)" (Kreher and Stinson, 1999, p. 1). These include, for example, algorithms for finding a clique of a certain size in a given graph. Many interesting combinatorial problems are, unfortunately, NP-hard.

  • Related Bachelor Projects:
    • Belief State Compression 1
      Supervised by:
    • Belief State Compression 2
      Supervised by:

Computer graphics and animation

Computer graphics is one of attractive fields in the world of digital media. Building 3D worlds and animating 3D characters as well as soft and rigid objects provide a fascinating experience. In most of our projects, we will use the famous Maya software along with programming different tasks to achieve advanced results. The programming languages we will use are MEL (Maya Embedded Language) and C++.

Computer Vision

Computer Vision is a field in visual computing that deals with images to extract useful information from them. Different topics are tackled in this field. We will work on the following projects.

Constraint Handling Rules

Constraint reasoning finds more and more applications. The rule-based concurrent programming language Constraint Handling Rules (CHR) was introduced to ease the development of constraint solvers. Currently several CHR libraries exist in languages such as Prolog, Haskell and Java, worldwide more than 50 projects use CHR. CHR and dozens of its constraint solvers can be used online via the internet at WebCHR. Over time it has become apparent that CHR and its extensions are useful for implementing reasoning systems in general, including deduction and abduction, since techniques like forward and backward chaining, bottom-up and top-down evaluation, integrity constraints, tabulation can be easily implemented and combined.

  • Related Bachelor Projects:

Education

Educational reform is occurring throughout the world and one of the tenets of the reform is the introduction and integration of assistive technological means in educational systems. Motivated by the prospect of greater economic, social, educational and technological gains, both developing and developed countries, are bringing about this reform, with a clear focus on technology integration in education. This cluster looks at how developing countries can adopt, adapt, and apply the knowledge gained by countries that have already embarked on the technological assistive means integration bandwagon in their own educational systems. We aim at illustrating the ways in which technology mediated learning can be offered for all students with their different profiles in order to achieve a better learning experience and knowledge gain.

  • Related Bachelor Projects:
    • Pen-A-Boo: The acceptability of using Smart Pens and AR Books fo
      Supervised by:
    • Laß Oder Las: A Serious Game for Analyzing Reading Behavior for German Language Beginners Using Eye-
      Supervised by:
    • Cut Or Cute: A Serious Game for Teaching Children the Effect of Vowels and Double Constants on Prono
      Supervised by:
    • Eqraaly: A Serious Game for Tracking the Performance of Children in Reading Arabic Text Using Eye-Tr
      Supervised by:
    • The effectiveness of AR technology to teach basic programming concepts to children with autism
      Supervised by:
    • Screenless Computer Education: A study of possible techniques
      Supervised by:
    • Building an Arduino Simulator and Debugger
      Supervised by:
    • Interactive Robot for Teaching Computational Thinking (English Language)
      Supervised by:
    • Kodockly II: Enhancing kodockly with adaptive debugging
      Supervised by:
    • Kodockly II: Using Blocks to Teach Language
      Supervised by:
    • Kodockly II: Using Tangible Blocks to Teach Mathematics
      Supervised by:
    • A context-Aware Lecture Annotation Tool
      Supervised by:
    • Lego-Mindstorm Simulator
      Supervised by:
    • Collaborative Coding in a Robotic Visual Language
      Supervised by:
    • How close in my Answer? Studying Contexts of Answers for Dialectal Arabic
      Supervised by:

FOREST firesystem

The Internet of Things, IoT for short, is a bunch of tiny computers equipped with sensors and actuators and connected wirelessly to each other and to a bigger distant computer. The equipped tiny computers are actually called Real-time Embedded Systems-RTES. RTES interact with the physical world in real-time: they sense the environment using their sensors and they act on it using... You guessed it: their actuators! So how would this become handy for wildfires you might ask? Well, we could place RTES equipped with temperature and smoke sensors on trees. And on the other side of the country, litteratly, we can monitor the temperature and detect bizzare readings or the very first fumes of burning wood. Using localization techniques, the exact geographical position of the nascent fire would be identified. This is early forest fire detection. Once detected and identified, many fire control solutions CAN be implemented.

Formal Languages and Automata

The theory of formal languages and automata is one of the cornerstones of theoretical computer science and string processing algorithms. The theory comprises the Chomsky hierarchy, formal grammars, finite automata, pushdown automata, Turing machines, etc.

  • Related Bachelor Projects:
    • Automata Simulation Tool
      Supervised by:

Game with a Purpose and Crowd Sourcing

Game with a Purpose (GWAP) make use of the contributions submitted by players of different Web gaming platforms in order to collect data. The games are designed in such a way so that while users are playing, needed data is collected. Crowdsourcing is an emerging approach that involves humans in solving problems which, so far, have no pure algorithmic solutions. Both fields are increasingly gaining importance. They aim at making use of human intelligence.

  • Related Bachelor Projects:
    • AKidKode: A serious game to teach programming skills to children
      Supervised by:
    • Encouraging cooperative play for children with autism through a computer based intervention.
      Supervised by:
    • Introducing a Mixed/Augmented Reality Platform for Mathematics Concepts
      Supervised by:
    • Introducing a Mixed/Augmented Reality Platform for Physics Concepts
      Supervised by:
    • Digitization of Arabic for Accessibility: A Game-Based Approach
      Supervised by:
    • A Game with a Purpose for Teaching Artificial Intelligence and Machine Learning
      Supervised by:

Graph Analytics

Many big data applications in social networks, governments, business and science need to be managed and analyzed the huge amounts of graph data. Specific systems to manage and to analyze such graph data meet a number of challenging requirements including the basic performance, the resource utilization, the scalability, and various performance overheads. Graph systems include graph database systems, distributed graph processing systems such as Google Pregel, Giraph, GPS, GraphLab, etc. and graph dataflow approaches based on Apache Spark, Flink, and Storm.

Graphics and Visualization

While computer graphics is the field that is concerned with using utilizing input description to produce artificial digital images, visualization is the visual representations of abstract data. We will work on different graphics and visualization tools.

Industry 4.0

With industry 4.0 as a fourth technological advancement wave, sensors , machines and different parts of a factory would be connected beyond even a single enterprise. Such cyber-physical systems can interact with each other using different protocols. They can use the data to predict failure, adapt to changes, … etc.With Industry 4.0, it would be possible to gather and analyze data across different machines and places. This is expected to increase productivity and even modify the profile of the workforce (Rüßmann, Michael, et al.).

  • Related Bachelor Projects:
    • Digital Twinning: Simulating and Controlling a Robotic Arm for I
      Supervised by:
    • Digital Twinning: Simulating and Controlling Robots’ Movements for Industry 4.0
      Supervised by:
    • Digital Twinning: Virtual Twin for Simulating and Controlling Robots’ Movements for Industry 4.0
      Supervised by:
    • Mixed/Virtual reality Aided Training for Industry 4.0
      Supervised by:

Innovation in the Internet

The last decade has witnessed significant strides in improving Internet performance. The development of new standards such as HTTP 2.0 and the new HTTP 3.0, new transport layer standards such as QUIC, are now widely used by many websites, including the major players such as Facebook, Google, and Amazon. Nevertheless, there are still many possible venues for improvement. New directions explore machine learning techniques to reduce delays and limit congestion. Content-based routing has also been proposed as a promising solution. Additionally, privacy has now become a major concern worldwide. Accordingly, innovation in the Internet can no longer ignore this important challenge. Web browsers such as TOR provide some privacy preservation assurances, at the expense of speed. This cluster also addresses privacy preservation in the Internet.

  • Related Bachelor Projects:
    • Managing Network Congestion with Deep Reinforcement Learning - C
      Supervised by:
    • Managing Network Congestion with Deep Reinforcement Learning - S
      Supervised by:
    • Reinventing Internet Routing for Privacy- Can we ditch the sourc
      Supervised by:
    • Reinventing Internet Routing for Efficiency - Content-based Rout
      Supervised by:
    • Creating a Simulated Dark Web Environment
      Supervised by:
    • Reinventing Internet Routing for Privacy - Hidden Services
      Supervised by:

Intelligent Autonomous Robots Control

Reactive autonomous robots navigating in real-world unstructured environments (i.e. environments that have not been specifically engineered for the robot) must be able to operate under conditions of imprecision and uncertainty present in such environments. The choice of adequate methods to model and deal with such uncertainties is crucial for robots operations in unstructured environments.

  • Related Bachelor Projects:
    • Adaptive pyramid implementation using mean values in full 2D space
      Supervised by:

Interaction Design and Formal Evaluation

Interaction design is a broad area within Human Computer Interaction (HCI) whose goal is creating fluid interfaces and natural ways of interacting with software. Whether it being a new menu, a set of new gestures to perform a basic task such as drag and drop, or a completely unconventional way to interact with a computing platform or the physical world, interaction designer and researchers are guided by theoretical findings coming mostly from psychology, physiology and few other disciplines. The thesis projects under this topic are all HCI and interaction design focused. Because HCI is offered in the 9th semester, most of you will be unaware of the basic concepts of HCI. However, if you take one of the below projects, I will be giving you a crash 1 day course in the beginning of the semester to help you start thinking like an HCIer!

  • Related Bachelor Projects:
    • Effect of Language on Code Understanding
      Supervised by:
    • MultiLingual Parser
      Supervised by:
    • Enhancing Electronics Lab Experience
      Supervised by:
    • Enhancing Doctor-Patient Communication via AR
      Supervised by:
    • Holographic Screen Design and Interaction Techniques
      Supervised by:
    • Study on ADHD Monitoring System
      Supervised by:
    • Intelligent Engineering-Drawing Table
      Supervised by:
    • Intelligent Haptic Desk
      Supervised by:
    • Intelligent Pillow
      Supervised by:
    • Mid-Air Typing on Cell Phone Keyboard
      Supervised by:
    • Multi-Device Tangible Screen
      Supervised by:
    • Speed-Oriented Interaction
      Supervised by:
    • Survey and Taxonomy of Motor Rules
      Supervised by:
    • Survey and Taxonomy of Input Hardware & Techniques
      Supervised by:
    • Survey and Taxonomy of Output Hardware & Techniques
      Supervised by:
    • Directional Voice Input
      Supervised by:
    • Crowd sourced Quality Tool
      Supervised by:
    • Creating Animations from EEG Signals I
      Supervised by:
    • Creating Animations from EEG Signals II
      Supervised by:
    • Creating Animations from Script I
      Supervised by:
    • Creating Animations from Script II
      Supervised by:
    • Hierarchy Navigation Using Gestures
      Supervised by:

Internet of Things (IoT)

Intelligence in the Internet of Things

  • Related Bachelor Projects:
    • An interface to assist children with non-verbal autism in communication.
      Supervised by:

ITS

Intelligent Transportation System –ITS- uses advanced technologies of electronics, communications, embedded computers, control actuators and detecting sensors in all kinds of transportation systems in order to improve safety, efficiency and service, traffic situation, and -most recently- pollution. this is done through capturing and transmitting real-time information.

Java SNePS

SNePS is one of the oldest, yet developing, systems for knowledge representation and reasoning in AI. Akin to a programming language, SNePS is a flexible system for building and manipulating propositional semantic networks. A propositional semantic network is a labeled directed graph where nodes represent entities (possibly propositions) and arcs represent structural relations among them. A SNePS network is thought of as the "mind" of an artificial, intelligent agent. The basic SNePS system is a network management system concerned with building and finding nodes. On top of that, there are several sub-systems for reasoning, acting, belief-revision, and natural language understanding and generation. The official version of SNePS is implemented in Common Lisp. Over the past seven years, we have implemented local Java versions. We are now ready to launch the third version Java SNePS. In the course of the development, we shall take extra care of implementation efficiency and proper software engineering practices.

  • Related Bachelor Projects:
    • Java SNePS Task 1
      Supervised by:
    • Java SNePS Task 2
      Supervised by:
    • Java SNePS Task 3
      Supervised by:
    • Java SNePS Task 4
      Supervised by:
    • Java SNePS Task 5
      Supervised by:
    • Java SNePS Task 6
      Supervised by:
    • Java SNePS Task 7
      Supervised by:
    • Java SNePS Task 8
      Supervised by:
    • Java SNePS Task 9
      Supervised by:
    • Java SNePS Task 10
      Supervised by:

Java SNePS

SNePS is one of the oldest, yet developing, systems for knowledge representation and reasoning in AI. Akin to a programming language, SNePS is a flexible system for building and manipulating propositional semantic networks. A propositional semantic network is a labeled directed graph where nodes represent entities (possibly propositions) and arcs represent structural relations among them. The basic SNePS system is a network management system concerned with building and finding nodes. On top of that, there are several sub-systems for reasoning, acting, belief-revision, and natural language understanding and generation. The official version of SNePS is implemented in Common Lisp. Over the past seven years, we have implemented local Java versions. We are now ready to launch the third version Java SNePS. In the course of the development, we shall take extra care of implementation efficiency and proper software engineering practices.

  • Related Bachelor Projects:
    • Java SNePS Task 1
      Supervised by:
    • Java SNePS Task 2
      Supervised by:
    • Java SNePS Task 4
      Supervised by:
    • Java SNePS Task 5
      Supervised by:
    • Java SNePS Task 3
      Supervised by:
    • Java SNePS Task 7
      Supervised by:

Machine Learning for Medical Diagnosis and Analysis

Machine learning has recently gained the attention of the medical research community for automated diagnosis of various diseases in addition to the analysis of medical data. Advances in machine learning and data analytics have enabled the analysis of patients’ data to aid physicians in their diagnosis as well as helping patients.

  • Related Bachelor Projects:
    • Predicting Epileptic Seizures from Recorded Heart Signals
      Supervised by:
    • Mortality Prediction in Intensive Care Units
      Supervised by:
    • Sleep Stage Prediction from Heart Rate and Motion Data
      Supervised by:
    • Classification of Neurodegenerative Diseases using Gait Analysis
      Supervised by:
    • Automated Leukemia Detection using Deep Learning Techniques
      Supervised by:
    • Automated Heart Arrythmia Recognition using Deep Learning Techniques
      Supervised by:

Massively Scalable Apps

This topic focus on the architecture, design and implementation of massively scalable applications such as those of social networks. Related areas include data structures and algorithms for scalability, distributed databases, data stores, distributed caches, farms, message queues, asynchronous I/O, load balancing, scalability, architectural design patterns, cloud architecture, design, and implementation.

  • Related Bachelor Projects:
    • Adding Observability/Migration to a Micro-blogging Platform
      Supervised by:
    • Enhancing Clustering in an open Source Database Engine
      Supervised by:
    • Ethereum Dapp Framework
      Supervised by:
    • Egy CryptoCurrency
      Supervised by:

Metrics and Measurements

Metrics are measure of quantifiable or countable characteristics. Metrics are important for many reasons, including measuring software performance, planning work items, measuring productivity, and many other uses.

Mobile Apps

For these projects, the student should have a basic knowledge of Android or iOS programming.

  • Related Bachelor Projects:
    • Bus ticket reservation mobile app
      Supervised by:
    • Online shopping mobile app
      Supervised by:
    • Jobs mobile app
      Supervised by:

Multimedia Analytics

Day to day vastly more media data are produced. It is became a part of everyday culture. Many people capture and edit visual and other multimedia data using simple devices. They share it through social networks, which have become the new leading tool mass communication. Multimedia data content are not yet automatically discovered. Most of tools for multimedia indexing and retrieval rely on metadata generated automatically at media production time or randomly by the users. In these projects we aim to address this problem by studying recent contribution in content-based indexing and retrieval, automatic or interactive multimedia data annotation, image and video forensics, large-scale image analysis, video semantic search engines, video synopsis, etc…

  • Related Bachelor Projects:
    • Verification of Video Originality in Online News
      Supervised by:
    • Image and Video Compression based on Multiresolution Analysis
      Supervised by:
    • Deep Learning Model for Quran Verses Recognition
      Supervised by:
    • Scene Classification in Performing Arts Videos
      Supervised by:
    • Music instrument classification using CNN
      Supervised by:
    • Music Scripting Using Machine Learning Techniques
      Supervised by:
    • Low Power High Accuracy Weather Station for Agriculture using LoRaWAN End Devices on The Things Netw
      Supervised by:
    • Mobile-based Grocery Products Query System
      Supervised by:
    • Real-Time Communication Interpreter System for Deaf and Dumb people using a Machine Learning Approac
      Supervised by:

Multimedia Networking

Multimedia streaming over data networks is extremely important in entertainment industry, communication systems, business applications and social networks. The audio and video streaming protocols keep evolving trying to meet new challenges and fulfill ever growing demand. This group of projects allows students, who are interested in this area, to learn more about up-to-date technologies and practice the implementation of streaming systems for a variety of goals. Although, these goals can be achieved using yesterday’s technologies, students will be asked to use today’s technologies.

Natural Language Processing for Code-Mixed (Arabic-English) Data

Within each country, native speakers tend lately to use code-mixing (CM) or Code-Switcing (CS) and they exchange between their own dialect and other languages. CM refers to the behaviour of mixing more than one language in the same context or conversation. For instance, In Egypt people use code-mixing for switching between Arabic and English in the same sentence e.g., “ana rayeh el lab”. It is a common behaviour that happens in written and spoken languages. Nowadays, Natural Language Processing (NLP) is extensively discussed and researched. It has been used in large fields like machine translation, speech recognition, and text processing. One of the most important tasks for NLP is Named-Entity Recognition (NER). NER systems were found to be extremely significant for various tasks in NLP as Information Retrieval and Question Answering tasks. NER is the task of identifying named entities and classifying them into different types or categories such as persons, locations and organizations. Much work has been conducted in this task for some major languages, such as English. However, there is less work done for Arabic specially Arabic-English CM data. One of the main problems in applying NER on Arabic-English CS data is the lack of annotated data. Thus, we collected and annotated the first corpus for CM Arabic-English data for NER tasks. The following projects are part of a collaboration with the Institute for Natural Language Processing (IMS), University of Stuttgart consisting of multiple workshops and opportunities taking place both at the GUC and Stuttgart University, where students working on these projects can attend.

  • Related Bachelor Projects:
    • Data Augmentation Technique for Code-Mixed Named Entity Recognition Corpus using Deep Learning
      Supervised by:
    • Classical Data Augmentation Techniques for Code-Mixed Named Entity Recognition Corpus
      Supervised by:
    • Transfer Learning for Code-Mixed Named Entity Recognition Technique
      Supervised by:
    • Code-Mixed Contextual Embedding
      Supervised by:
    • Language Identification on the Subword-level for Code-Switching
      Supervised by:

On-device Machine Intelligence

The current paradigm for using machine learning models on a device is to train a model in the cloud and perform inference using the trained model on the device. However, with the increasing number of smart devices and improved hardware, there is interest in performing model training on the device. This interest also stems from the need to train models on individual devices to preserve the privacy of data hosted on these devices and send them to the cloud. Functions like conversational understanding and image recognition can be performed locally on smartphones via learning partial machine learning models, and then the models are pooled and averaged to provide a holistic view of the task without exposing the underlying data.

  • Related Bachelor Projects:
    • Building Federated Language Models from Smartphone Text
      Supervised by:
    • Deep Learning for Mobile Sensed Data: Activity Recognition
      Supervised by:
    • Federated Learning for Mobile Sensed Data: Activity Recognition
      Supervised by:
    • Deep Learning for Mobile Sensed Data: Image Tagging
      Supervised by:
    • Federated Learning for Mobile Sensed Data: Image Tagging
      Supervised by:
    • Deep Learning-based Time Series Forecasting for Mobile Sensor Da
      Supervised by:
    • Optimized Deep Learning Networks for Mobile Devices
      Supervised by:

Reinforcement Learning Applications

Reinforcement learning refers to goal-oriented algorithms, which learn how to attain a complex objective (goal) or maximize along a particular dimension over many steps; for example, maximize the points won in a game over many moves. They can start from a blank slate, and under the right conditions they achieve superhuman performance. Like a child incentivized by spankings and candy, these algorithms are penalized when they make the wrong decisions and rewarded when they make the right ones – this is reinforcement.

  • Related Bachelor Projects:
    • Deep Reinforcement Learning for Job Recommendations
      Supervised by:
    • Deep Reinforcement Learning for Talent Recommendation
      Supervised by:
    • Analysis of Reinforcement Learning Security Vulnerabilities
      Supervised by:
    • Deep Reinforcement Learning for Path Planning
      Supervised by:
    • Deep Reinforcement Learning for Text Chatting
      Supervised by:

Remote Sensing

Remote Sensing Images refer to image acquisition using satellite- or aircraft-based sensor technologies to detect and classify objects on Earth. It has many applications in many fields such as security, traffic managements, civil engineering, environmental monitoring and others.

  • Related Bachelor Projects:
    • Road Network Extraction Using Deep Learning Model
      Supervised by:
    • Deep Learning Models for Semantic Segmentation of Hyperspectral Image
      Supervised by:
    • On the Application of Deep Learning for Road Network Extraction
      Supervised by:

Simulation of Prosthetic Vision

Visual prostheses hold hope of vision restoration for millions with retinal degenerative diseases. The main principal of visual prosthesis is to bypass the damaged part of the human visual pathway and electrically stimulate subsequent functional parts. Such electrical stimulation is supposed to deliver electrical pulses to the visual pathway that mimic what would have been received if the damaged part was intact. There are different types of visual prostheses; some of them have already been approved to be used to restore vision to blind patients, while others are under development. The most successful type is the retinal implant in which stimulating electrodes are used to replace damaged photoreceptors; converting light into electric pulses to stimulate the undamaged parts of the retina. Despite the success of many of these devices, the image perceived by blind patients who are using these devices is very primitive. Perceived images are only black and white with very limited resolution. Given that our understanding of the perceived image is mainly based on the verbal description of the implanted patients, there is currently multiple research efforts to simulate how such perceived image looks like. Such simulation studies are expected to enhance our understanding of prosthetic vision and, thus, enhance the quality of the perceived image.

  • Related Bachelor Projects:
    • Moving Object Recognition Enhancement for Visual Prostheses
      Supervised by:
    • Enhancing Arabic Character Recognition for Visual Prostheses
      Supervised by:
    • Enhancing Face Recognition for Visual Prostheses
      Supervised by:

Smart Home Applications using Brain-Computer Interfaces

Almost all motor disabled people as tetraplegia patients, lose their normal muscular control. Consequently, they are entirely locked into their bodies. Though they are conscious, yet they have no means to move or communicate with their surrounding environment. This may consequently lead to a greater dependence on others and expose those patients to social exclusion. One way to help those patients is to interpret their intentions into commands. Brain-Computer Interface (BCI) acts as an output channel to help those people to communicate with the external environment. They can benefit from BCI technology in communication and control, motor substitution, entertainment, and motor recovery. BCI attains brain signals and translates them into commands enabling those patients to operate devices that help them overcome their disabilities. BCI can be used to help disabled people to interact with and manage his accessible living environment using brain signals.

  • Related Bachelor Projects:
    • Wheelchair Control using Brain-Computer Interface
      Supervised by:
    • Home Appliances Control using Brain-Computer Interface
      Supervised by:
    • Home Robots Control using Brain-Computer Interface
      Supervised by:

Smart office

A “smart office” is an office space equipped with IoT devices. It can be seen as an ecosystem relying on those IoT devices that: monitor, control, and manage various operations and working conditions. Among the many benefits of a smart office: Energy savings, efficient business operations, comfortable working environment, better employee productivity, increased workplace safety…

  • Related Bachelor Projects:
    • [Lab]A Smart Secured system for visitor management and access co
      Supervised by:
    • AR for the hearing impaired
      Supervised by:
    • Smart sensitive planter system 2.0
      Supervised by:

Smart Spaces and the Internet of Things

The Internet of Things has ushered in an era where objects around us can interact with us to create smarter spaces and ecosystems. This cluster investigates several domains where smart spaces can be created. The projects in this cluster range from smart cities, to smart campuses, to smart offices. The projects integrate knowledge from the fields of sensor networks, artificial intelligence, and embedded systems to address issues that we face in our daily lives.

  • Related Bachelor Projects:
    • Realizing an Internet of Things for the Human Character (Character IoT)
      Supervised by:
    • Federated Knowledge in the IoT
      Supervised by:
    • Ubiquitous Ctrl+C/Ctrl+V - Collaboration between devices on a new scale
      Supervised by:

Smart-X

This is a series of projects embedding intelligence for different applications and purposes

Sound and Acoustics

Acoustics is the field of science that is concerned with the generation, propagation and reception of mechanical waves and vibrations in gases, liquids, and solids including sound, ultrasound and infrasound. Understanding acoustic phenomena is indispensable for digital media engineers who wish to work in the recording industry.