We do Deep-Learning.
Nov 14, 2017 • Lyndon Maydwell
Previously there have been a lot of skeletons overlaid on top of video at the office, but with Andy Gelme’s recent creation of office data-topics and the publication of pose data, it’s easier than ever to hook in and create something a little more interesting. Here you can see pose information being consumed by an Elm application written by Noon to render skeletons with Obama’s head. (What do you mean ‘Why?’)
What are some fun ideas for how this tech could be used?
- Saving prep and processing-time for animation motion-capture
- Gait analysis / Pose-ID
- Game input / Simon-Says style games
- Crowd mood detection
- Scene construction
- Image warping
- South-Park style cartoon creator / photo-shopping by moving
- Robotic control! (Dancing?)
The differentiator with products like the Kinect is that pose-estimation works purely off video/image data, rather than requiring any augmented sensing. This commodifies the capability into any web-cam connected to the internet.
Come along to the Deep-Dance Smackdown to see some of this technology in action.