region_generator.py

import os
import skimage
import matplotlib.pyplot as plt
import selectivesearch
import numpy as np
from PIL import Image


dataset_path = "FlickrLogos-v2/"
bbox_path = "FlickrLogos-v2/classes/masks/"
images_path = "FlickrLogos-v2/classes/jpg/"
DIRACTORY_VAL= "regions/"


def create_folders():
	if not os.path.exists(DIRACTORY_VAL):
		os.makedirs(DIRACTORY_VAL)	
		

# -- Function that ....
# --   
# -- 
def read_data(data_file):
	samples = {}
	with open(dataset_path + data_file, 'r') as f:		
		for line in f:			
			label, filename = line.split(",")
			#if (label != "no-logo"):
			filename = filename.rstrip()
			sample_dic = {}
			sample_dic['label'] = label
			samples[filename] = sample_dic	
			                             
	return samples
			  
# -- Function that ....
# --   
# -- 
def read_boundingboxes(bbox_file):
	boundingbox = {}
	with open(bbox_file,'r') as f: 
		header = True
		for line in f:
			if not header:
				x,y,w,h = line.split(" ")	
				boundingbox['x'] = int(x)
				boundingbox['y'] = int(y)
				boundingbox['w'] = int(w)
				boundingbox['h'] = int(h.rstrip())
			else:
				header=False
	return boundingbox

# -- Function that generates annotated region proposals from an image using selective search
# --   
# --   
def generate_region_proposals(img_name,img_propery):
	
	label = img_propery['label']
	image_filename = images_path + label + "/" + img_name
	img = Image.open(image_filename)
	img_arr = np.asarray(img)
	img_lbl, regions = selectivesearch.selective_search(img_arr, scale=500, sigma=0.9, min_size=10)	
	candidates = set()

	# create a txt file where it saves generated regions for that image
	path_region = DIRACTORY_VAL + img_name[:-4] + ".txt"
	file = open(path_region,"w")
	file.write("x y width height" + "\n")
	
	for r in regions:
		generated_sample_reg = {}
		# excluding same rectangle (with different segments)
		if r['rect'] in candidates:
			continue
		# excluding regions smaller than 2000 pixels
		if r['size'] < 2000:
			continue
		# distorted rects
		x, y, w, h = r['rect']
		if w == 0 or h == 0:
			continue
		if w / h > 5 or h / w > 5:
			continue	
		candidates.add(r['rect'])
		#label = annotate_region(img_propery,r['rect'])	
		#save regions of a picture on a txt file
		x, y, w, h = r['rect']
		txt = str(x) + " " + str(y) + " " + str(w) + " " + str(h)  + "\n"
		file.write(txt)
	file.close() 
	

# -- Function that annotates an image region with a label (the logo name or 'no logo')
# -- by comparing it with the ground truth bounding box of the logo 
# --   Returns: string representing the label of the input region
def annotate_region(prop, region):
	label = prop['label']
	box = prop['box']
	iou  = IOU(box,region)
	# annotate
	if (iou < 0.4):
		label = 'no-logo'
	return label  


def IOU(box,region):
	# get corner coordinates of bounding box
	xA1 = box['x']
	yA1 = box['y']
	xA2 = xA1+box['w']
	yA2 = yA1+box['h']
	
	# get corner coordinates of region
	xB1, yB1, wB, hB = region
	xB2 = xB1 + wB
	yB2 = yB1 + hB
	
	# return 0 if no intersection at all
	if (xA1>xB2 or xA2<xB1 or yA1>yB2 or yA2<yB1):
		return 0.0
	
	# determine the (x, y)-coordinates of the intersection rectangle
	xA = max(xA1, xB1)
	yA = max(yA1, yB1)
	xB = min(xA2, xB2)
	yB = min(yA2, yB2)
	
	# compute the area of intersection rectangle
	interArea = (xB - xA + 1) * (yB - yA + 1)
	
	# compute the area of both the prediction and ground-truth
	# rectangles
	boxAArea = (xA2 - xA1 + 1) * (yA2 - yA1 + 1)
	boxBArea = (xB2 - xB1 + 1) * (yB2 - yB1 + 1)
	
	# compute the intersection over union by taking the intersection
	# area and dividing it by the sum of prediction + ground-truth
	# areas - the interesection area. +0.001 in case of division by zero.  
	iou = interArea / float(boxAArea + boxBArea - interArea + 0.001)
	return iou


	

# -- Function that ....
# -- samples  
# -- set : {1 for trainset.txt , 2 for trainvalset.txt }
def generate_training_data(samples):
	i=0
	for img_name,img_property in samples.items(): 
		generate_region_proposals(img_name,img_property)
		i+=1
		print(i)
	

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


#-- Create output_python/ directory (folder)
create_folders() 

#-- Read the dataset images
training_data = read_data("testset.txt")
generate_training_data(training_data)