Louis E. King from University of Michigan was talking yesterday about methods of adding time-synced meta video to aid searches of video repositories. This data including things like:
- thumbnail of frames
- text to speech
- face recognition
- OCR
The demo of the server-based system that managed all this was very impressive and I thought it was blowing my mind, but then my mind was truly blown:
In addition to the sources of metadata I listed above, the system can also take backchannel communication like the activity of a live question tool or twitter and timesync it to the video. So, you can watch the video and view the synced tweets as well as use the tweets to help find sections of video related to particular topic. Cool!
Leave a comment