Logo der Universität Wien

Open Topics for Bachelor Theses

If you are looking for a Bachelor Thesis topic please register for the course offered for the Bachelor Thesis and contact us for the thesis topics currently available in the semester. In general, all the topics listed below are also available for Bachelor Thesis projects. Of course, the topic will be limited in effort and scientific claim to meet the requirements of a Bachelor Thesis. If you are interested do not hesitate to contact us; send e-mail to Prof. Wolfgang Klas or contact a member of the research group. Before contacting us, PLEASE read the Recommendations & Guidelines for Bachelor Thesis available herehttp://cs.univie.ac.at/students/der-weg-durchs-studium/bachelorarbeit/

Open Topics for Master Theses and Practical Courses (PR, P1, P2)

In the following some of the open topics in the area of Multimedia Information Systems are listed. If you are interested and if you have an idea on a project do not hesitate to contact us; send e-mail to Prof. Wolfgang Klas or contact a member of the research group.

In general, topics in the area of Multimedia Information Systems technologies include:

  • analyze, manage, store, create and compose, semantically enrich & play back multimedia content;
  • semantically smart multimedia systems;
  • security.

Possible application domains include:

  • Content Authoring and Management Systems
  • Web Content Management
  • Robotic and IoT Applications
  • Blockchain Technologies and Applications
  • Interactive Display Systems

Section (A) below lists topics that can be chosen in the course of a PR Praktikum, but are in principles also available for a master thesis (usually expanded and more advanced).

Section (B) below lists topics that are intended to be chosen for a master thesis.

 

 

(A) Topics for Practical Courses (PR, P1, P2) in the Master Degree Programs

FactCheck - Entity Recognition based on dedicated Cognitive Computing Services

FactCheck is a framework for the detection and resolution of conflicting structured data on the Web. The FactCheck framework is the result of ongoing research at our research group. One of the central building blocks is the identification of objects represented by means of structured data (encoded as Microformat or JSON-LD data) on different web pages, but constituing the same conceptual real world entity or artefact. This problem is also known as entity recognition.

Goal of the project is to design and implement suitable, alternative solutions for entity recognition based on at least two cognitive computing services available as cloud services from e.g., IBM, Microsoft, Google, supporting text data, figure/image data, video and audio data. Results of the project are to be demonstrated by a running demo application.

Technologies: web services, semantic web technologies, LOD, Microformat, JSON-LD, dedicated cloud computing services

Contact: Wolfgang Klas wolfgang.klas@univie.ac.at

FactCheck - Precision Metrics

FactCheck is a framework for the detection and resolution of conflicting structured data on the Web. The FactCheck framework is the result of ongoing research at our research group. One of the central building blocks is the context-dependent comparison of structured data of various representations of one and the same real world object or artefact. The comparison is guided by so called precision metrics which is a flexible and sophisticated technique for logically comparing structured data values. Precision metrics consist of logical predicates used to evaluate the comparison of structured data. Goal of the project is to design and implement an appropriate model for the representation of precision metrics, the construction of such precision metrics as well as the application of the metrics for evaluating the comparison of data values. Various precision metrics should be defined and compared using a test dataset of 900.000 entities. Results of the project are to be demonstrated by a running demo application.

Technologies: web services, semantic web technologies, LOD, Microformat, JSON-LD

Provided to the students: existing implementation of framework, test dataset

Contact: Wolfgang Klas wolfgang.klas@univie.ac.at

Demo of Blockchain application using Proof-of-Authority

The CS Faculty operates the recently founded BlockchainSci-Lab. For students the lab offers an environment to get familiar and work with state of the art systems and platforms to learn about blockchain technology and to design and implement blockchain applications by participating in dedicated projects.

The goal of this project is the implementation of a demo application which illustrates the concept of proof-of-authority (in place of the very often used "proof-of-work" as, e.g., used in the Bitcoin Blockchain). For example, a possible application could be the implementation of the four-eyes principle (Vier-Augen-Prinzip) for officially approving documents by making use of two signers acting as proof-of-authorities. There are many other possible application scenarios feasible. It should be well-chosen in order to illustrate the general principle of proof-of-authority. It may be based on a generic, configurable implementation to show different variations of the proof-of-authority concept, e.g., 1 signer, 2 signers, N signers.
The demo application has to be realized such that a short demonstration movie can be recorded, that will be published on the Lab's website.

Provided to the students: Optionally, Blockchain-IT infrastructure with initial configuration of proof-of-authority based on Ethereum Project.

Contact: Wolfgang Klas   wolfgang.klas@univie.ac.at

"Studienleistungs & Prüfungspass" based on MultiChain Blockchain technology

The CS Faculty operates the recently founded BlockchainSci-Lab. For students the lab offers an environment to get familiar and work with state of the art systems and platforms to learn about blockchain technology and to design and implement blockchain applications by participating in dedicated projects.

The goal of this project is to implement an application for a digital "Studienleistungs & Prüfungspass" (study performance & examination pass) based on blockchain technology. The pass will record the individual, required study achievments (like milestones, tests, etc.) during a course, the final grading of a course, and the collection of gradings of courses during the entire study (like a "Sammelzeugnis" currently used by the university). There are various stakeholders in this scenario: the students and the lectures of courses. The implementation has to be realized based on MultiChain Blockchain technology, which provides managed permissions and allows for millions of assets on a blockchain, structured asset data, multiple key-value, time series or identity databases on a blockchain.  MultiChain Blockchain technology is one of the most promising implementations for private, managed blockchain systems.

Technology: MultiChain Blockchain Infrastructure, Version 1.0.1, on Linux of Windows, or on Cloud Infrastrcture.

Provided to the students: Optionally, IT-Infrastructure of the BlockchainScie-Lab, virtual machine

Contact: Wolfgang Klas   wolfgang.klas@univie.ac.at

Using Blockchain to track images - Web application – Python, flask

The CS Faculty operates the recently founded BlockchainSci-Lab. For students the lab offers an environment to get familiar and work with state of the art systems and platforms to learn about blockchain technology and to design and implement blockchain applications by participating in dedicated projects.

News agencies suffer from the problem of dealing with faked images that are processed or changed for many reasons. Fortunately, Blockchain technology is one of the most creative ideas that can generate linked data blocks, which are secured using cryptography. Thus, news agencies can exchange images (and data about images) more safely. The goal is to build a smart system that can help reporters (in general any photographers) to add their images into a Blockchain to exchange images and ensure the quality of daily news. 

Goals of the project:

  • Record/Capture image
  • Design a Blockchain model
  • Find suitable cryptography approaches  

Based on your findings implement a framework, which is able to 

  • upload an image
  • add the image into a Blockchain
  • look for similar images on the Web (using Google APIs)
  • create a user interface to see changes of the image's Blockchain
  • report the status of the image  

Your program must be able to report and track your image well to ensure that it is under your control.

Contact: Nour Jnoub   nour.jnoub@univie.ac.at 

Recognize the name of a song and retrieve similar songs - Android or Python

We hear songs mostly everywhere and it is common to hear a nice song but you do not know, what is the title to search and hear it later again. Although, there are many similar apps that can help you to get the title of a song, they may not provide you with a list of similar songs that you may like as well. 

Goals of the project: 

  • Capture a song
  • Design  a database for songs
  • Find a suitable  recognition and recommendation model

Based on your findings implement a framework, which is able to  

  • capture a song
  • retrieve the title of the song from a database based (as an audio signal)
  • suggest a list of recommended songs
  • create a user interface to add songs to your playlist and favourites

Your application must be able to  suggest songs in an acceptable quality

Contact: Nour Jnoub   nour.jnoub@univie.ac.at 

Sorting images based on face expressions - python application

Do you want to see your happy, sad, or funny memories ? Do you have an unmanageable number of photos? Do you know how to sort them automatically? This topic focuses on sorting different types of images and profiling them automatically based on faces and their expressions. 
Goals of the project:

  • Design a software architecture (UML)
  • Find suitable face detection approach
  • Find suitable face expression recognition approach 

Based on your previous findings implement a framework, which is able to  

  • Sort images based on different face categories (emotions) 
  • Sort images based on a reference face expression image
  • Create a user friendly interface
  • Report the status and the quality of your sorting quality (score) 

Your program must be able to perform well and be able to avoid false positives as good as possible.

Contact: Nour Jnoub   nour.jnoub@univie.ac.at 

Online Media Tracking Considering Context, Sematic and Causalities

The development of trend scouting tools, leveraging signals from online media, has recently been gaining increasing attention. These tools help to extract trends (influential topics) in order to proactively address problematic issues and situations which occur instead of reacting later. In addition to the basic goals of such systems which continuously monitor unstructured online media for identifying trending topics, forecasting future dynamic of topics are even more important goal.   
Long short-term memory (LSTM) is a recurrent neural network (RNN) architecture that remembers values over arbitrary intervals. RNNs allow forward and backward connections between neurons. An LSTM is well-suited to classify, process, and predict time serises given time lags of unknown size and duration between important events. Nevertheless, for accurate forecasting of trending topics, two modelling factors are important: 

  • Integration of correlations and causalities among topics.
  • Integration of contextual and semantical changes of topics.  

The main goal of this project is to investigate mentioned factors and conducting an experimental evaluation for development of accurate LSTM model for forecasting future dynamic of topics. 
 
Technologies: Python, Machine Learning libraries (Tensorflow or Keras) 

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

Natural Language Retrieval using Elasticsearch

Natural language search, opposed to keyword-based search, is phrasing queries as you would ask them if you were talking to someone. These questions can be typed into or spoken with a search engine (such as asking a question to a digital assistant like Siri or Cortana). Natural language search is an attempt to break down a query into the most important terms, remove unnecessary connecting words like “what”, “or”, “the”, and so on. So if you wanted to know how high the Millennium Tower is, a keyword-based search query for that information might be “Millennium Tower height”. But if you were searching using natural language, you would phrase your query as, “How high is the Millennium Tower?”  

The main goal of this project is to develop a system, which: 

  • uses NLP methods to prepare natural language questions,
  • extracts main information/entities of the questions,
  • uses Elasticsearch to retrieve relevant results to the questions.  
    Technologies: Natural Language Processing, Information Retrieval

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

Active learning on Mobile 

There are large numbers of mobile health applications which collect biological information such as physical activities or mood. Many of these applications leverage collected data to predict mood and behaviour of users. Nevertheless, users of these systems have different behaviours and habits which change in different context, time and environment, therefore a static training set for creating activities and behavioral predictors is not accurate and effective enough. The main goal of this project is to develop a framework which considering paradigm of active learning enables the system to adapt itself according to the potential changes in users’ behaviour over time.  

Technologies: Android, Machine Learning  

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

Identification of Useful Sensor for Mood Predictor 

The promise of “Mood Predictor” using mobile sensing has long been championed. However, different issues (such as missing data, uncertainty) limit the usage of these solutions in practice. Nevertheless, users’ activities and interactions can be also sensed from wearables and their on-line social interactions. Therefore, exploring different combinations of various sensors enable the development of the next generation of “Mood Predictor”.  
The main goals of this project are to:  

  • Develop a mood predictor app, which leverages and combines data from different sensors from various sources (wearables, online social media, mobile sensing).
  • Identify the influence of various sensors for different types of persons.

Technologies: Android, Machine Learning  

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

Identification of Dynamic of Trending Topics Considering Context and Semantic

Tracking trending topics from online media has recently gained increasing attention. Most of the available solutions assess the influence of topics only considering frequency of topics over time without considering context and semantic changes of topics. For example for the trending topic, "Sebastian Kurz", assessing the influence of the topic over time considering only frequency scores of the topic does not show high changes, as the frequency scores of this topic over the time are always high. However, the topic has high contextual and semantical changes as "Sebastian Kurz" took over different political roles such as " State Secretary for Integration" or "Chairman of the Party".   
This work uses a neural network model called Word2Vec on each aggregated group of documents related to different time periods or regions, which helps to obtain syntactical, contextual and semantic information of each topics.  Develop an application: 

  • to identify semantical and contextual changes of trending topics from temporal corpus of online documents, and
  • to visualise semantical and contextual changes.

Technologies: Natural Language Processing, Visualisation 

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

Causality Inference for Media Trend Scouting 

The development of media trend scouting tools has recently been gaining increasing attention. These tools help to extract trends (influential topics) in order to proactively address problematic issues and situations which occur instead of reacting later. In addition to the basic goals of such systems which continuously monitor unstructured media data from the open Web, understanding occurrences and identifying correlations and causalities are even more important goals.  

The main goal of this project is to develop a system, which by receiving a set of temporal topics and trends (extracted from online news) 

  • assesses the correlation between topics and trends using semantic correlation methods (such as co-similarity of topics in a temporal context and content),
  • assesses the correlation between topics and trends using temporal causality methods (such as Granger causality methods), and
  • recommends causal retationships for trends considering semantic and temporal correlations.  

Technologies: Natural Language Processing, Data Analytic  

Contact: Elaheh Momeni-Ortner elaheh.momeni-ortner@univie.ac.at

[P1-DR1-17] Motion Capture through Marker Tracking

Create an application which captures, analyses and saves the movements of the upper human body. The application tracks 3 or more optical/fiducial markers attached to the upper body of a person with a camera. Captured data, results of an analysis of the motion (speed, turn rate,…) and optionally additional data (user data, description,…) can be saved by the user if wanted. 

Technologies: CLion (https://www.jetbrains.com/clion/), C++, OpenCV 

Required skills: Basic C++ Knowledge, Possibility to print, Laptop with Camera or PC+webcam 
  
Contact: Daniela Ramsauer   daniela.ramsauer@univie.ac.at

[P1-DR2-17] Depth Images Preprocessing

Depth images are often used for tasks like skeleton tracking or object recognition. Due to hardware and cost limitations depth images are of low quality and holes occur in the data. Find and implement three different depth image enhancement  algorithms. Compare the results and document the differences. 

Technologies: CLion (https://www.jetbrains.com/clion/), C++  

Provided to the student: Depth images of low quality 

Required skills: Basic C++ Knowledge, Jetbrains Students Account (For educational use only) (https://account.jetbrains.com/login)  

Contact: Daniela Ramsauer  daniela.ramsauer@univie.ac.at

[P1-DR3-17] Skeleton Tracking - Depth Images

Implement and evaluate the  algorithm for Human Skeleton Extraction of Depth Images Using the Polygon Evolution described in [1] with data from [2].
 
Technologies: CLion (https://www.jetbrains.com/clion/), C++  

Required: Basic C++ Knowledge, Jetbrains Students Account (For educational use only) (https://account.jetbrains.com/login 

References: 
[1]
Du, Huan and Wang, Jian and Zhong, Xue-xia and He, Ying and Mei, Lin", Editor "Duffy, Vincent G.", Human Skeleton Extraction of Depth Images Using the Polygon Evolution", in "Digital Human Modeling. Applications in Health, Safety, Ergonomics and Risk Management: 5th International Conference, DHM 2014, Held as Part of HCI International 2014, Heraklion, Crete, Greece, June 22-27, 2014. Proceedings", 2014, Springer International Publishing,
ISBN 978-3-319-07725-3,
doi 10.1007/978-3-319-07725-3_2 

[2] 
University of Coimbra 3-D Motion Database, 
Data available at mrl.isr.uc.pt/experimentaldata/public/uc-3d/  

Contact: Daniela Ramsauer daniela.ramsauer@univie.ac.at

[P1-DR4-17] Selflearning Track-Your-Back App - Inertial Sensor

In rehabilitation it is often important to analyse the movements of a patient. This is usually done by doctors or physiotherapists. Inertial sensors nowadays create the possibility for automatic testing and analysis of movements. The goal is to create an Android app, which is able to detect if a user is sitting and then analysis the movements of the back. 
For this purpose the user wears a phone close to his/her body (e.g., in a pocket trouser) and the MetamotionR inertial sensor is attached to the upper back. 
If the app detects that the user is sitting, the app starts collecting orientation data from MetamotionR inertial sensor. Next, movement features like ‘lean-forward’, ‘lean-back’, and the like should be used to save the inertial data in the most efficient way (less storage needed – maximum information gain). The collected information should be analysed to give feedback to the user about his upper back movement habits.  
Furthermore the app should contain a possibility for the user to add a back variable (good, little back pain, intensive back pain). This variable is assigned to the previous inertial data collected and used to warn the user if pain causing movement patterns happen.  

Technologies: Android Studio, Android Java, Metawear SDK (https://mbientlab.com/developers/) 

Provided to the student: inertial sensor MetamotionR 

Required skills: Android device with Bluetooth and inertial sensors (Gyroscope and  Accelerometer)

Contact: Daniela Ramsauer daniela.ramsauer@univie.ac.at

 

... additional, new topics will become available soon.

(B) Topics of Master Theses

Please check the listing below for possible topics for a master thesis. In principle, you may also chose from the topics listed in Section (A) above. Those topics are available for a master thesis as well, but usually in a more expanded or advanced form.

 

... additional, new topics will become available in near future. In the case of Master Theses topics you may also contact a researcher of the MIS group to find out more about possible topics.

Contact us
Faculty of Computer Science
University of Vienna

Währinger Straße 29
A-1090 Vienna