Changing shirt colours would be a fairly straightforward segmentation problem. Applying patterns accurately would require some kind of 3d reconstruction, which

Changing shirt colours would be a fairly straightforward segmentation problem. Applying patterns accurately would require some kind of 3d reconstruction, which is obviously harder. I like the gesture controlled UI idea. Understanding sign language is probably several steps too far - I think it relies heavily on facial expressions to give the gestures context, and it almost certainly isn't a straight mapping from gestures to words.. Will hunt around for literature though, as even understanding a small subset would be fascinating and I'd love to learn to sign.