Deep learning and dance synthesis

Through this master's project on motion capture, the student will acquire advanced knowledge of motion capture data and deep learning. 

With motion capture technology, we can obtain precise quantitative representations of dance movement. With recent advances in machine learning and recurrent neural networks, we may train computers to imitate human behaviour. The approach has been demonstrated in automated image captioning, synthesised speech, and generative music.

The student choosing this master's project, will explore deep learning applied to motion capture recordings of dance movement, enabling automatically generated dance patterns, either in an avatar or a robot.

Suggested qualifications:

Machine learning / Deep learning (e.g. INF5860, INF4490),  Motion capture (MUS2006), Biokinematics.

Publisert 31. mai 2017 13:49 - Sist endret 31. mai 2017 13:51

Veileder(e)

Omfang (studiepoeng)

60