Hand-pose estimation is an important problem for Man-Machine Interface systems. In this work, we have proposed a hand-pose estimation system that can estimation the palm orientations and the 20 angles of the joints of five fingers from a single camera image. The complete system integrates several CNN models for hand identification and segmentation, location of 2D key points, and solving inverse kinenmatics of the joint angles. |
|
Reference:
|
Demo videos |