Multimodal lifelong learning of human nonverbal behavior

The project goal is to develop ML solutions for complex video based human behavior modeling in real world environment and the transition of results to practice in healthcare. The model should be able to analyze and represent the heterogeneous, dynamic, nonuniform encoding process of expressive behaviors across clinical conditions and extrapolate that knowledge across contexts and populations.

Prerequisite knowledge

Strong background in machine learning/computer vision/human behavior modeling, with specialization in one or more of the following areas: supervised/unsupervised/self-supervised learning, large vision models, multimodal foundation models, transfer learning, Research experience in one or more of the following areas: multimodal human behavior modeling, affective computing, AI for healthcare, multimodal machine learning.