NewTest.py

From IPRE Wiki

from pyrobot.brain.conx import *

def shrinking(pic):
    
    X = getWidth(pic)
    Y = getHeight(pic)

    # Input the shrink factor and computer size of new image
    F = 8
    newx = X/F
    newy = Y/F

    # create the new image
    newPic = makePicture(newx, newy)

    for x in range(1, newx):
        for y in range(1, newy):
            setPixel(newPic, x, y, getPixel(pic, x*F, y*F))

    return newPic


def locateObject(picture):
    #This is what I'm using to check to see if the neural network returns the
    #correct results
    tot_x = 0
    count = 0

    for pixel in getPixels(picture):
        r, g, b = getRGB(pixel)
        if g == 255:
            tot_x = tot_x + int(getX(pixel))
            count = count + 1
    if count > 0:
        return tot_x/count
    else:
        return -1


def inputPictures():
        # take picture and locate the object
        pic = takePicture()
        show(pic, "normal")

        nPic = shrinking(pic)
        picList = []

        show(nPic)

        #I could have just left this as a grayscale image but I already wrote the
        #code changing it to black and white. I realize I made more work for myself
        for pixel in getPixels(nPic):
            r, g, b = getRGB(pixel)
            if g > 200 and r < 130:
                setRGB(pixel, (255, 255, 255))
                picList.append(1)
            else:
                setRGB(pixel, (0, 0, 0))
                picList.append(0)
        show(nPic, "changed")

        #Object location is not fed into the neural network. This prints out the
        #location so that I can check it with the neural network's results
        objectLocation = locateObject(nPic)
        print objectLocation

        if objectLocation < 0:
            print "Not in Picture"
        elif objectLocation <= 11:
            print "Left"
        elif objectLocation <= 22:
            print "Center"
        else:
            print "Right"
            
        return picList #returns the list of 0's and 1's for the input
    

def main():
    #FORMAT python

    # create the network
    n = Network()

    # add layers in the order they will be connected
    n.addLayer('input',738)      # The input layer has # nodes
    n.addLayer('output',3)     # The output layer has # node
    n.connect('input','output')  # The input layer is connected to the output layer

    # provide training patterns (inputs and outputs)
    theInput = []
    for i in range (3):
        theInput.append(inputPictures())
        wait(5)
   
    n.setInputs([theInput[0], theInput[1], theInput[2]])
    n.setOutputs([[0.0, 1.0, 0.0], [1.0, 0.0, 0.0], [0.0, 0.0, 1.0]])

    # set learning parameters
    n.setEpsilon(0.5)
    n.setTolerance(0.2)
    n.setReportRate(1)

    # learn
    n.saveWeightsToFile("before.wts")
    n.train()
    n.saveWeightsToFile("after.wts")

    # verify learning
    n.setLearning(0)
    n.setInteractive(1)
    n.sweep()