Teaching & theses

In the fall term 2026, we are teaching:

• Human-Computer Interaction 2026 bachelor course
• Mixed Reality 2026 master course

Available student thesis projects

7 projects are available (via 5 master’s theses, 4 semester theses, and 3 bachelor’s theses). The list below is up-to-date.

Please submit your CV and transcripts through SiROP. In case of questions, please contact us via SiROP.

• Learning Interaction-Aware Robot Hand Retargeting from monocular RGB videos

  Human hand demonstrations provide useful training data for dexterous robots. However, directly mapping human hand motion to a robot hand often produces unstable contact, object drops, or incorrect object motion.
  In this Master’s thesis, you will train a temporal model to improve robot hand controls using corrections generated by an interaction-based teacher. The teacher uses the object mesh during training, but the learned model will only use observed hand and object motion during inference. The project will test whether this approach can improve manipulation success while avoiding costly physics optimization for every new demonstration.

  more...
  
  • contact: Berken Demirel
  • master's thesis

  • Dexterous manipulation
  • Interaction
  • hand retargeting
  • imitation learning
  • temporal models
  • human demonstrations
  • object manipulation

• Multi-Modal Spatial Reasoning with Human Motion

  Multi-modal language models still struggle with the 3D world — judging directions, taking another viewpoint, estimating distances — and prior work traces most of these failures to egocentric–allocentric frame transformation, not perception. Humans learn this by moving through space. This thesis will test whether full-body human motion can supply that signal. The plan is to extract a spatial-reasoning dataset coupled to real human motion, train an MLLM on it, and measure whether the learned ability transfers to existing benchmarks.

  more...
  
  • contact: Dominik Hollidt
  • semester project
  • bachelor's thesis
  • master's thesis

  • spatial reasoning
  • MLLM
  • VLM
  • egocentric video
  • human motion
  • embodied AI
  • proprioception
  • instruction tuning
  • benchmark

• Motion Capture Data Collection for Robotics Retargeting and Sparse IMU-Based Human Motion Estimation

  This project will collect a realistic motion-capture dataset for sparse wearable IMUs, targeting applications in human pose estimation and robotics motion retargeting. Unlike existing laboratory datasets, it will capture real-world effects such as sensor noise, drift, misalignment, and imperfect placement using affordable commodity hardware.

  more...
  
  • contact: Dominik Hollidt
  • semester project
  • bachelor's thesis

  • motion capture
  • robotics
  • retargeting
  • motion
  • IMU
  • inertial sensing
  • pose estimation
  • deep learning
  • machine learning
  • computer vision
  • benchmark
  • dataset

• Hand-object manipulation through generated sensor layouts on 3D surfaces for robotic grasping

  We will develop a generative geometry-processing method for creating surface-conforming sensor layouts on discrete 3D object surfaces. Using an existing fabrication pipeline, the resulting instrumented objects will capture spatial contact and force during human manipulation, providing ground-truth supervision for robotic grasp learning, imitation from human demonstrations, and egocentric vision.

  more...
  
  • contact: Christian Holz
  • semester project
  • master's thesis

  • hand-object manipulation
  • robotic grasping
  • geometry processing
  • surface meshes
  • contact sensing
  • force sensing
  • egocentric perception

• Motion Capture Data Collection for Benchmarking Deep Learning-Based Pose Estimation from Sparse IMUs

  Deep learning-based human pose estimation from a small number of body-worn inertial measurement units (IMUs) is crucial for practical AR/VR and healthcare applications. However, most existing datasets are collected under laboratory conditions with high-end hardware and carefully calibrated, perfectly placed sensors. As a result, they implicitly assume near-ideal sensor behavior and mounting, which does not reflect the noisy, misaligned, and drift-prone data typically produced by affordable commodity IMU devices in real-world use.
  This project focuses on recording a new motion dataset that explicitly captures these real-world imperfections. The goals include:
  - Collecting synchronized IMU and reference motion capture data across a diverse set of activities
  - Running a user study to collect high-quality, systematically recorded IMU data from non-specialized IMU devices across a diverse group of participants
  - Preprocessing and packaging the recordings for seamless integration with an existing benchmarking platform

  more...
  
  • contact: Dominik Hollidt
  • semester project
  • bachelor's thesis

  • pose estimation
  • deep learning
  • machine learning
  • computer vision
  • benchmark
  • dataset

• Supervised Contrastive Learning for Cross-Sensor Tactile Representation Learning in Robotics

  Tactile sensors help robots understand contact, force, slip, texture, and objects. However, different sensor types produce very different data, which makes it difficult to train models that work across sensors.
  In this Master’s thesis, you will develop a learning method that combines self-supervised learning with sparse labels such as force, motion direction, and indenter type. The goal is to learn useful tactile representations that transfer to new sensors and physical conditions.

  more...
  
  • contact: Berken Demirel
  • master's thesis

  • Tactile sensing
  • representation learning
  • self-supervised learning
  • supervised contrastive learning
  • robotics.

• Tactile Representations in Robotics via Self-Supervised and Optimal-Transport-Based Learning

  We will develop and validate a self-supervised learning pipeline for tactile representation learning across different tactile datasets and sensor types. Using existing public tactile datasets, we will train tactile encoders on images and align the learned feature spaces using Optimal Transport. The goal is to evaluate whether optimal transport can improve cross-dataset and cross-sensor transfer without requiring exact paired tactile samples.

  more...
  
  • contact: Berken Demirel
  • master's thesis

  • self-supervised learning
  • optimal transport
  • representation learning
  • tactile

Current thesis students

• Yuchen Qin master's UZH Computer Science
• Tong Su master's UZH Computer Science
• Le Qian master's UZH Computer Science
• Qichen Zhao master's UZH, Computer Science
• Nicolas Baeriswyl master's ETH Computer Science, D-INFK
• Qi Ma master's ETH Computer Science, D-INFK
• Ines Kast semester ETH Robotics, D-MAVT, RSC
• Andri Bernhardsgrütter semester ETH Data Science, D-INFK
• Monica De Alvaro Mena semester ETH Data Science, D-INFK
• Damian Herculano semester ETH Computer Science, D-INFK
• Tingting Xu semester UZH Computer Science
• Gujer Sascha bachelor's ETH Data Science, D-INFK

Past Courses and Seminars

• Mobile Health and Activity Monitoring master course
  • 2026
  • 2025
  • 2024
  • 2023
  • 2022
• Human-Computer Interaction bachelor course
  • 2025
  • 2024
  • 2023
  • 2022
  • 2021
  • 2020
  • 2019
• Mixed Reality master course
  • 2025
  • 2024
  • 2023
• Ubiquitous Computing master course
  • 2021
  • 2020
• Adaptive User Interfaces through Machine Learning bachelor seminar
  • 2026
  • 2025
• Egocentric Computer Vision: Perception and Understanding from the First-Person View master seminar
  • 2026
• Reinforcement Learning for Understanding and Modeling Human Behavior master seminar
  • 2026
  • 2025
• Egocentric Perception for Human-World Understanding master seminar
  • 2025
• Advanced Topics in Mixed Reality bachelor seminar
  • 2024
  • 2023
  • 2022
• Advanced Topics in Technical Human-Computer Interaction bachelor seminar
  • 2021
  • 2020