Anna Deichler

Master thesis projects
I have master thesis projects available, in connection to recent work on MM-Conv: A Multi-Modal Conversational Dataset for Virtual Humans. The projects are ML-oriented, focusing on multi-modal data (vision/speech/motion). Contact me at deichler@kth.se for more details.
=====================================================
Focus Area
In my thesis, I focus on enabling embodied agents to engage in human-like interactions by integrating multiple data modalities—text, audio, and motion, with a particular emphasis on motion generation in situated contexts.
My recent work explores reinforcement learning [1] and generative models [2] for gesture generation, with a focus on referential communication. In future work, I am interested in the combination and trade-off of these methods. I am also interested in how recent advances in language modeling can be leveraged in interactive embodied agents and the use of pre-trained models within modular architectures. Furthermore, I am interested in the use of VR for training interactive agents. This research is pivotal for applications where agents must communicate effectively within physically situated environments using both verbal and non-verbal modes.

Deichler, Anna, et al. "Learning to generate pointing gestures in situated embodied conversational agents."Frontiers in Robotics and AI 10 (2023): 1110534.
Deichler, Anna, et al. "Diffusion-based co-speech gesture generation using joint text and audio representation."Proceedings of the 25th International Conference on Multimodal Interaction. 2023.