instance_map.py

# import the necessary packages
from skimage.exposure import rescale_intensity
from skimage.segmentation import slic, mark_boundaries
from skimage.util import img_as_float
from skimage import io
from scipy import misc
import numpy as np
import argparse
from PIL import Image
import glob
from tqdm import tqdm
import os
from joblib import Parallel, delayed

# Directories
atri_dir = 'attribute_512p/'
image_dir = 'images_512p/'
segmentation_dir = 'seg_512p/'
output_dir = 'instance_map_no_border/'

if not os.path.exists(output_dir):
    os.makedirs(output_dir)

file_name_arr = []
for file in glob.glob(atri_dir+'*.png'):
	temp = file.split('/')[-1].split('_')
	file_name = temp[0]+'_'+temp[1]
	if file_name not in file_name_arr:
	    file_name_arr.append(file_name)

def create_instance_map(family):
	# Create a zero filled base image
	# Load original image
	image = misc.imread(image_dir+family+'.png')
	instance_map = np.zeros(image.shape[:2], dtype=int)
	segments = slic(img_as_float(image), n_segments=1000,
	slic_zero=True, compactness=1, sigma=2)

	for i, file in enumerate(glob.glob(atri_dir+family+'*.png')):
		# Read Mask
		mask = misc.imread(file)
		type_file = file.split('/')[-1].split('_')[3]
		if i ==0:
			segmentation = misc.imread(segmentation_dir+family+'_segmentation.png', flatten=True)
			last_lesion = 2000
			last_background = 1000
			for v in np.unique(segments):
				union = segmentation[segments==v]
				if float(len(union[union > 0])) / float(len(union)) > 0.5:
					instance_map[segments == v] = last_lesion
					last_lesion += 1
				else:
					instance_map[segments == v] = last_background
					last_background += 1

		if type_file == 'pigment': # 3
			last = 3000
		elif type_file == 'negative': # 4
			last = 4000
		elif type_file.startswith('streaks'): # 5
			last = 5000
		elif type_file == 'milia': # 6
			last = 6000
		elif type_file.startswith('globules'): #7
			last = 7000
		else:
			print('ERROR: Invalid File Found!!!!')
		# For each superpixel in the selected mask, update the pixel values incrementing the value for each superpixel found.
		for v in np.unique(segments):
			union = mask[segments == v]
			if float(len(union[union > 0])) / float(len(union)) > 0.5:
				instance_map[segments == v] = last
				last += 1
	instance_map = instance_map.astype(np.uint32)
	im = Image.fromarray(instance_map)
	im.save(output_dir+family+'_instance.png')

	
results = Parallel(n_jobs=8)(delayed(create_instance_map)(family) for family in tqdm(file_name_arr))