The Institute of Computational Perception continuously supervises student projects and theses at all levels (BSc, MSc, PhD) in areas like AI, machine learning (including, of course, Deep Neural Networks and Deep Reinforcement Learning), intelligent audio and music processing, multimedia, recommendation systems, biometrics, etc. (see below for details).
Interested and motivated students are invited to contact us at any time during the semester.
The following theses and practical courses are offered in every winter and summer semester:
If you have further questions or if you are interested in one of our topics below (or if you want to propose your own topic) do not hesitate to get in touch with our Faculty Staff .
This category comprises specific project topics that we are particularly interested in at a particular time - for instance, because they are part of a larger research project running at our institute. Sometimes, we may be able to offer financial support for these. Sometimes, this list may also be empty ...
Current focus 3 (SS 2020): Using machine learning for analysis and
organization of museum collection (see Project Descriptions)
If you are interested in any of these, contact Gerhard Widmer for more information.
Our project topics are structred into thematic areas, which also relate to the specific research experience and interests of our team members. Each thematic category lists a corresponding contact person.
Machine Learning and Artificial Intelligence
Much of our research applies AI and Machine Learning (see further below), but we are also interested in developing, improving or understanding Deep Learning or Statistical Learning/Modeling in general: How does it work? When does it fail? What can we do when data is scarce? Projects may involve re-implementing methods from the literature, systematic experiments with designed data sets, attempts at qualitative or performance improvements, etc. These projects generally require a background in Machine Learning.
Intelligent Audio Processing
Our institute is heavily involved in research on Intelligent Audio Processing. In this context, we offer a variety of projects involving audio data and machine learning, and many different learning scenarios ((semi-)supervised, transfer learning, learning with weak labels, learning interpretable models, ...), e.g. participation in the annual DCASE http://dcase.community/challenge.
Intelligent Music Processing
As one of the leading research labs in the fields of Sound and Music Computing and Music Information Retrieval (MIR), we offer diverse topics related to AI & Music. Examples include real-time tempo and rhythm recognition, instrument detection in audio, music transcription, optical music recognition, music identification, live music tracking, but also intelligent music interfaces or ML-based models of generating music. You should have an interest, and a bit of background, in music, and at least basic knowledge of Machine Learning. See here for some examples of what we do.
Multimedia Data Mining
Analyzing and mining of large-scale datasets including multimedia content and user-generated data has become a highly active research area, often positioned at the intersection between computer science and social sciences. At CP, we offer a variety of topics related to this area. Examples include semantic analysis of text, music, video, and metadata, e.g., employing methods of unsupervised and supervised learning, content-based feature extraction, and statistical analysis. In addition, we are interested in analyzing and mining online social networks (e.g., investigating their network structure, connectivity, homophily, etc.).
Multimedia Recommender Systems
We offer topics on all aspects of user modeling and recommender systems. Our focus is on psychologically and sociologically inspired user models, for instance, models of personality, mood, cognition, or culture. Elaborating algorithms to predict such user characteristics (for instance from user-generated data) is of particular importance for user-aware systems.
In this vein, topics can also involve developing algorithms for user-aware and context-aware recommender systems, devising methods to increase transparency of recommendations, and fusing several approaches (e.g., collaborative filtering and content-based filtering) to create hybrid recommender systems.
The main goal of digital image processing is to extract useful information about the objects contained in an images from an input image of potentially low quality. The steps necessary to this end (restoration, geometry correction, enhancement, segmentation, feature extraction, classification) are therefore of major interest for us at CP. Additionally we are very interested in image copyright protection (digital image watermarking)
Decades ago, biometric identification methods were only used in special areas of application (forensics, access to high-security areas), but nowadays biometric identification is almost everywhere (just think about biometric passports). For us at CP, the whole variety of biometric traits (finger, face, iris, retina, hand, ear, voice, signature, gait, etc.) is of interest.
Cryptography („secret writing“) is a science that provides essential methods for secure communication. To be more specific, cryptographic methods can ensure confidentiality, integrity, authentication and nonrepudiation when a sender transmits a message to a recipient. At CP we are interested in all relevant cryptographic concepts, algorithms and protocols to ensure these properties for secure communication.
Natural language processing
Understanding and processing natural language has been a challenging area of AI research for decades. Recent advancements with deep learning (DL) have brought a promising perspective but also many new topics and questions. In CP, we offer topics on fundamental research on deep learning models in NLP, various aspects of DL models such as bias and energy efficiency, and end-user NLP applications.
Machine Learning for Symbolic Music Processing
Symbolic Music Processing deals with all aspects of symbolic representations of music (e.g., machine readable musical scores). Topics include modeling and generating expressive music performance, score-to-performance alignment and analysis of musical structure. You should have an interest in music (reading sheet music would be very helpful), and some basic knowledge of Machine Learning.