MSc thesis projects

In the Socially Perceptive Computing Lab, MSc theses work (TU Delft students) in the domain of automated human social and affective behaviour analysis with a special focus on non-verbal behaviour analysis in dyadic or group and team settings. Projects vary in scope from more domain specific machine learning tasks to more computational social science tasks. Depending on the project, there is also the opportunity to work in collaboration with social scientists.

Below is an indication of the types of themes that can be covered in a thesis:

• Domain specific machine learning is present in all projects to a greater or lesser degree depending on the preference of the student in consultation with the supervisor.  Paradigms include but are not limited to transfer learning, domain adaptation, deep learning, multiple instance learning, active learning, social network analysis, and/or unsupervised clustering.
  
  
• Basic individual behaviour modelling topics for modality specific or multimodal problems such as (but not limited to) multimodal multi-sensor fusion, semantic segmentation, social action recognition, gesture recognition, pose or posture estimation, or person Re-ID, tracking, and detection.
  
  
• Estimation of socially related phenomena or behaviour modelling such as (but not limited to) team cohesion, social attraction, or conversational involvement, emergence, multimodal synchrony such as between speech and gesture.
  
  
• Estimation of affect related phenomena such as (but not limited to) estimating individual affect, personality, or social quality of experience.
  
  
• Other topics related to the themes of the group such as (but not limited to) user feedback, smart experience sampling,  mobile and/or wearable computing.