Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Explore behaviour correlation in natural conversations
We want to perform human behaviour analysis in a natural conversational setting as well as develop solutions for interesting tasks, such as speech-based gesture synthesis. Provided dataset contains multiple modalities, include hand pose, body pose, eye gaze, facial expression and speech.
Keywords: Deep learning, natural conversation, multiple modalities
In natural conversations, we express ourselves through various channels (e.g. hand/body pose, facial expression etc.). However, available modelling only focuses on one or two aspects. In this project, we want to further analysis human behaviour in natural conversation settings using a recently collected dataset from the group, and develop methods for downstream tasks, such as speech-based hand pose generation. Available codebase would facilitate the development of the project. Dataset contains multi-view RGB input and sound recordings. Ground truth data includes hand poses, body poses as well as gaze directions.
Feel free to contact us if you find the project interesting, and we are open to new suggestion and ideas.
In natural conversations, we express ourselves through various channels (e.g. hand/body pose, facial expression etc.). However, available modelling only focuses on one or two aspects. In this project, we want to further analysis human behaviour in natural conversation settings using a recently collected dataset from the group, and develop methods for downstream tasks, such as speech-based hand pose generation. Available codebase would facilitate the development of the project. Dataset contains multi-view RGB input and sound recordings. Ground truth data includes hand poses, body poses as well as gaze directions.
Feel free to contact us if you find the project interesting, and we are open to new suggestion and ideas.
Explore the potential of a newly collected dataset in a natural conversational setting.
Explore the potential of a newly collected dataset in a natural conversational setting.
Dr. Xucong Zhang xucong.zhang@inf.ethz.ch
Dr. Xi Wang xi.wang@inf.ethz.ch
Dr. Xucong Zhang xucong.zhang@inf.ethz.ch Dr. Xi Wang xi.wang@inf.ethz.ch