l1 = list1.copy()

l2 = list2.copy()

for l in l1:

if l in l2:

l2.remove(l)

else:

return False

'''

def __init__(self, json_data):

self.json_data = json_data

self.categories = {}

self.coco = {} #the main dictionary

for im in self.json_data["images"]:

self.coco[im["id"]] = {"url": im["flickr_url"]} #saving url for each img

for cat in self.json_data["categories"]: #makes categories dictionary, that contain all possible objects.

self.categories[cat["id"]] = {"name": cat["name"], "supercategory": cat["supercategory"]}

anns_per_image = {}

print ("Loading Coco Metadata")

for ann in tqdm(self.json_data["annotations"]): #for each image, it creates a list with annotated objects

if ann["image_id"] not in anns_per_image:

anns_per_image[ann["image_id"]] = {"label_ids": [ann["category_id"]],

"label_texts": [self.categories[ann["category_id"]]["name"]],

"bboxes": [ann["bbox"]]}

else:

anns_per_image[ann["image_id"]]["label_ids"].append(ann["category_id"])

anns_per_image[ann["image_id"]]["label_texts"].append(self.categories[ann["category_id"]]["name"])

anns_per_image[ann["image_id"]]["bboxes"].append(ann["bbox"])

# we join categories with the main dict of imgs

for im_id in list(self.coco):

if im_id in anns_per_image:

self.coco[im_id]["label_ids"] = anns_per_image[im_id]["label_ids"]

self.coco[im_id]["label_texts"] = anns_per_image[im_id]["label_texts"]

self.coco[im_id]["bboxes"] = anns_per_image[im_id]["bboxes"]

else:

self.coco.pop(im_id, None) # αν δεν έχουμε annotations σβήνουμε το key αυτό

def prune_dataset(self, labels):

#receives a label list of objects and drops from dataset

#all images that don't contain at least once each item in that label list.

image_ids = list(self.coco.keys())

for image_id in image_ids:

if not is_sublist(labels, self.coco[image_id]["label_texts"]):

self.coco.pop(image_id, None)

def join(self, other):

#joins coco dictionaries of two different datasets!

#if A, B are datasets, this is AUB

self.coco = {**self.coco, **other.coco}

def random_image(self):

#returns a random image from the dataset

random_int = randint(0, len(self.coco))

return self.coco[list(self.coco.keys())[random_int]]

def find_uniques(self):

#define unique variable

self.unique_objects = set()

#iterate through dataset to find unique objects

for id in self.coco:

self.unique_objects.update(set(self.coco[id]['label_texts']))

#makes list from set and sorts alphabetically

self.unique_objects = list(self.unique_objects)

self.unique_objects.sort()

#change to dictionary for a more helpful form...

self.unique_objects = {object: num for num, object in enumerate(self.unique_objects)}

def bag_of_words(self):

#calculates bag of words embedding for each image in the dataset.

for id in self.coco:

self.coco[id]['bow'] = [0 for _ in self.unique_objects]

for object in self.coco[id]['label_texts']:

index = self.unique_objects[object]

self.coco[id]['bow'][index] += 1

def clean_urls(self):

#cleans the data from images that simply dont have the url anymore!

#note this takes quite a while if you have bad internet connection!

for id in tqdm([*self.coco.keys()]):

url = self.coco[id]['url']

response = requests.get(url)

if response.status_code != 200: