RuntimeWarning: overflow encountered in ubyte_scalars

Question

I'm new to Python and this is my first ever thing I've scripted and I'm just wondering what I can do to remove this warning:

Warning (from warnings module):
  File "C:\Users\Luri\Desktop\Bot Stuff\ImageSaver.py", line 76
    currentdiff=abs(anread[w,h])-abs(bnread[w,h])
RuntimeWarning: overflow encountered in ubyte_scalars

I've tried Googling the answer and nothing that was clear to me came up as far as fixing this.

I'm trying to write a program that will compare a continuously updating image that is taken from a rectangle around my cursor with a reference image that I'm searching for.

Then depending on what region the cursor is in relative to the target image, it will adjust accordingly.

Thank you for any help you can give!

-J

Code is below:

import os
import sys
import time
import Image
import ImageGrab
import win32api
import numpy, scipy

def mousePos():
#---------------------------------------------------------
#User Settings:
  SaveDirectory=r'C:\Users\Luri\Desktop\Bot Stuff'
  ImageEditorPath=r'C:\WINDOWS\system32\mspaint.exe'
#Here is another example:
#ImageEditorPath=r'C:\Program Files\IrfanView\i_view32.exe'
#---------------------------------------------------------
  i,j = win32api.GetCursorPos()
  print 'Your Cusor Position is:', i,j
  time.sleep(1)
  size = 112, 58
#-------------------
#data is defined as | x0y0 = [0,0] = (xpos-56,ypos-29) | x0y1 = [0,1] = (xpos-56,ypos+29) | x1y1 = [1,1] = (xpos+56,ypos+29) | x1y0 = [1,0] = (xpos+56,ypos-29)
#Take In Image In Rectangle around cursor position to locate text of name
  pixeldiff=0
  currentdiff=0
  NQ1=193395
  NQ2=166330
  NQ3=171697
  NQ4=168734
  NAC=190253
  NBC=205430
  x0=i-56
  y0=j-29
  x1=i+56
  y1=j+29
  box=[x0, y0, x1, y1]
  img=ImageGrab.grab()
  saveas=os.path.join(SaveDirectory,'fullscreen.jpg')
  img.save(saveas)
  editorstring='""%s" "%s"'% (ImageEditorPath,saveas)
#Crop box around cursor
  cursorbox=img.crop(box)
  saveas=os.path.join(SaveDirectory,'cursorbox.jpg')
  cursorbox.save(saveas)
#Converts the given cursor rectangle to 8bit grayscale from RGB  
  out = cursorbox.convert("L")
  saveas=os.path.join(SaveDirectory,'lmodecurbox.jpg')
  out.save(saveas)
#Takes the converted grayscale picture and converts it to an array
  a=numpy.asarray(out)
  aarray=Image.fromarray(a)
  sizea = a.shape
#  print sizea
#  print a
  anread=a[:]
#Loads the reference image
  reference=Image.open("referencecold.png")
#Converts the given cursor rectangle to 8bit grayscale from RGB
  refout = reference.convert("L")
  saveas=os.path.join(SaveDirectory,'lmoderefbox.jpg')
  refout.save(saveas)
#Takes the converted grayscale picture and converts it to an array  
  b=numpy.asarray(refout)
  barray=Image.fromarray(b)
  sizeb = b.shape
#  print sizeb
#  print b
#  print size
  bnread=b[:]
#  print bnread
#Realized you can determine position based on this single quadrant
#Loop Quadrant 1 x0y1 to xmym
  for h in range(0,29):
    for w in range(0,55):
      #currentdiff=0
      currentdiff=abs(anread[w,h])-abs(bnread[w,h])
      pixeldiff=pixeldiff+currentdiff
#  print pixeldiff
#Test Above
  if pixeldiff<198559 and pixeldiff>190253:
  #Test Left
    if pixeldiff > 175000:
    #Move Above and Left
      print ('Go Up and Left')
    else:
    #Move Above Right
      print ('Go Up and Right')
  if pixeldiff>198559 and pixeldiff<205430:
    if pixeldiff < 185000:
    #Move Below and Left
      print ('Go Down and Left')
    else:
    #Move Below and Right
      print ('Go Down and Right')
"""
#Nominal Q1=193395 Variance low = 188408 Variance high = 203194
#Nominal Q2=166330 Variance low = 181116 Variance high = 199208
#Nominal Q3=171697 Variance low = 172279 Variance high = 201816
#Nominal Q4=168734 Variance low = 190644 Variance high = 191878
#Nominal Center = 198559
#Nominal Above Center = 190253
#Nominal Below Center = 205430
#Loop Quadrant 2 xmy1 to x1ym
  for h in range(0,29):
    for w in range(55,111):
      difference=abs(a(w,h)-b(w,h))
      currentdiff=abs(anread[w,h])-abs(bnread[w,h])
      pixeldiff=pixeldiff+currentdiff
#Loop Quadrant 3 x0ym to xmy0
  for h in range(29,57):
    for w in range(0,55):
      difference=abs(a(w,h)-b(w,h))
      currentdiff=abs(anread[w,h])-abs(bnread[w,h])
      pixeldiff=pixeldiff+currentdiff
#Loop Quadrant 4 xmym to x1y0
  for h in range(29,57):
    for w in range(55,111):
      difference=abs(a(w,h)-b(w,h))
      currentdiff=abs(anread[w,h])-abs(bnread[w,h])
      pixeldiff=pixeldiff+currentdiff
#Fine Nominal Values for Each quadrant pixeldiff
#Compare which is similar and then move cursor in center of that quadrant
"""

def main():
#  while True:
  mousePos()

if __name__ == "__main__":
  main()




#Compare image to constantly updating image of rectangle around cursor (maybe per second?) by searching for the quadrant with most similarity

#-------------------

#Based on comparison, move cursor to middle (x and y value) of matched quadrant by population of similar features and repeat

Negate

Answer 1

You are adding two uint8 values together resulting in an uint8 value. You need to transform your data types in the calculation. I suggest you try this:

pixeldiff = (int(ipxeldiff)+int(currentdiff))/2

This should work.

Edit: balanced bracketts

Scupper

The problem is already in the line before that. It needs to be converted to int() there. It's already too late in the pixeldiff line. – Spandex

Answer 2

I had a similar problem that I resolved by initializing my numpy arrays as an int64 datatype:

imAnchor = array(Image.open(imList[10]), dtype='int64')

Parfitt

that could increase the memory that the image occupy – Folger

Answer 3

I think your problem stems from this line:

pixeldiff=pixeldiff+currentdiff

Remember that pixels are normally saved in uint8 datatype, which is 0 to 255. So if you try to add two of them and it goes over 255, it will fail.

Do something like this:

pixeldiff = (pixeldiff+currentdiff)/2

You will still get the relational data, but it will be compressed into the right size of 0-255.

Hollah

The problem is in the line before that. It needs to be converted to int() there. It's already too late in the pixeldiff line. – Spandex

Answer 4

RuntimeWarning: overflow encountered in ubyte_scalars is because of datatype. so change the datatype of anread and bnread once: anread = anread.astype('int32') bnread = bnread.astype('int32')

Obsessive