https://kitzeller.github.io/accessible-space/
You can define gestures for WASD controls by making a gesture and then clicking the corresponding button. Each time you click the button it will add a training image to the KNNClassifer for that class. The more images, the more accurate predictions will be. Be sure to train a neutral image for when you want to stop movement (e.g. a neutral position)
- Tensorflow.js
- Babylon.js