ReMagicMirror

When people want to learn e.g. dances or martial arts moves, they can copy the motions from a video, or from a teacher in real life. However, videos can only be watched from a single viewpoint, which may make learning certain actions more difficult. And teachers aren't always available. So we made ReMagicMirror, which displays the action the user with the same orientation as the user's own body, making comparisons much simpler!

The system:

The process:

Publications

  1. Dayrit, F. L., Nakashima, Y., Sato, T., Yokoya, N. "ReMagicMirror: Action Learning Using Human Reenactment with the Mirror Metaphor." Submitted for MMM2017.

Augmented Reality Reenactment

We use a single RGB-D sensor to capture a person's movements, creating a reenactment - novel view synthesis using only a single viewpoint. The reenactment does not include any background; we fill this gap by using augmented reality to overlay it on top of a real environment.

Input:

Microsoft Kinect v2, RGB frames

Reenactment:

Microsoft Surface Pro 2

Publications

  1. Dayrit, F. L., Nakashima, Y., Sato, T., Yokoya, N. "Free-viewpoint AR human-motion reenactment based on a single RGB-D video stream." ICME 2014.
  2. Dayrit, F. L., Nakashima, Y., Sato, T., Yokoya, N. "Increasing pose comprehension through augmented reality reenactment." Multimedia Tools and Applications (2015).