process_images.py

from PIL import Image
import control_panel
import numpy as np
import os
from os import listdir
from os.path import isfile, join

size = control_panel.image_size
p = control_panel.image_pixels # number of features (pixels) in each image
X = np.zeros(shape=(1, p))  # initialise image matrix with zeros in one row
count = 1

def get_thumbs(File_Path, Out_Path):
    print("Processing photos in...", File_Path)
    photos = [f for f in listdir(File_Path) if isfile(join(File_Path, f))]
    for infile in photos:
        try:
            global count
            img = Image.open(File_Path + infile).convert("L")  # convert image to greyscale
            img = img.resize(size)  # resize image
            img.save(Out_Path + str(count) + ".png", "PNG") # save as new image in the Processed folder
            a = np.asarray(img).reshape(-1) # reshape each matrix of image's pixel alpha values to a vector
            global X
            X = np.vstack([X,a])    # add vector of each image's pixel alpha values to new row of Image matrix
            count += 1
        except IOError as e:
            print("cannot create thumbnail for" +e, infile)
    print("Finished processing ", File_Path)

def test_thumbs(File_Path, Out_Path):
    photos = [f for f in listdir(File_Path) if isfile(join(File_Path, f))]
    for infile in photos:
        outfile = os.path.splitext(infile)[0]
        if infile != outfile:
            try:
                img = Image.open(File_Path + infile).convert("L")  # convert image to greyscale
                img = img.resize(size)  # resize image
                img.save(Out_Path + outfile + ".png", "PNG") # save as new image in the Processed folder

            except IOError as e:
                print("cannot create thumbnail for" +e, infile)


get_thumbs("Data/Raw_Bex/", "Data/Processed/")
X = np.delete(X, (0), axis=0)   # delete the first row of Image matrix (the zeros row)
m = X.shape[0]

get_thumbs("Data/Raw_Peet/", "Data/Processed/")
n = (X.shape[0]) - m

get_thumbs("Data/Raw_Dad/", "Data/Processed/")
o = (X.shape[0]) - (n + m)

get_thumbs("Data/Raw_Joan/", "Data/Processed/")
p = (X.shape[0]) - (n + m + o)

bex = np.ones((m,1), dtype=np.int)     # create vector to append to front of image matrix, to label who is in photo (1 = bex)
peet = np.zeros((n,1), dtype=np.int)   # create vector to append to front of image matrix, to label who is in photo (0 = peet)
dad = np.ones((o,1), dtype=np.int) * 2  # create vector to append to front of image matrix, to label who is in photo (2 = dad)
joan = np.ones((o,1), dtype=np.int) * 3  # create vector to append to front of image matrix, to label who is in photo (3 = joan)


x = np.vstack([bex, peet])
x = np.vstack([x, dad])
x = np.vstack([x, joan])

X = np.hstack([x, X])
np.save('Data/image_data', X)  # save newly created image matrix to a data file, so can import to neural net


# Creates thumbnails of Test Images
#test_thumbs("Data/Test_Images/", "Data/Processed_Test/")