FocalSpace from Tangible Media Group on Vimeo.
http://tangible.media.mit.edu/project/focalspace/
Computer vision and background segmentation has been an interest of mine for a while. Having worked on these algorithms over the summer. I thought this project was interesting since it combines background segmentation with an intuitive (and brilliant) way of localizing the speaker.
The background segmentation is being supplemented by a kinect which provides a depth map that allows the extraction of the foreground. Additionally, a microphone array is used to localise the speaker. The kinect depth map along with the speaker localisation allows deriving a mask for the image to be blurred.
0 kommentarer:
Skicka en kommentar