The project goal was to design a working model of a human palm which replicates its motion and controls using a video feed from camera.
Poster - link
Using pose estimation, first we detect all the joints and segments and generate a heat map. Using these positions, we calculate the joint angles with change in position and these angles are transmitted to an arduino via serial transfer. The arduino, maps these angles to its corresponding motor angle. Now, as the motor rotates, the finger bends similar to a human finger in a direction depending upon the direction of rotation of the motor.
For in-depth explanation of the design and working, please read here
This is the Tensorflow implementation of Convolutional Pose Machines, one of the state-of-the-art models for 2D body and hand pose estimation.
Tracking support for single hand.
- Easy multi-stage graph construction
- Kalman filters for smooth pose estimation
- Simple self-tracking module
