Open thepooltable opened 6 years ago
I guess this should be done by Maths instead of a network ;)
For the circle; Check if the distance between the points is similar For the triangle; Check the linearity between 3 points
Both are convolutions; So if you really want to do it using networks; I would try: INPUT 1x10 CONVOLUTION 1x19 PERCEPTRON 1x19 OUTPUT 1x2
I'd like to use touch point detection, e.g 5 touch point: logging 5 pairs of clientX & clientY coordinates of the screen.
In a form of training 1000 times each, a set of 5 touch point in circle form, a set of 5 touch point in triangle form.
Am i correct to use, e.g 1 of 1000 diff testing values: "input":[382,201,519,216,451,172,449,222,524,253], "output":{"a":1, "b":0}}] - irregardless of the touch point placement.
if the 5 touch points are in a circle form, it will be able to recognise and show: a is more then 0.5 if the touch points are in triangle form, it should be: b is more then 0.5