From ed5a6f2cdbad1d8cc31901ae7f1e7e2386b910b3 Mon Sep 17 00:00:00 2001
From: bjoern <bjoern.deiseroth@aleph-alpha.com>
Date: Mon, 6 Nov 2023 16:40:11 +0000
Subject: [PATCH] more cleanup

---
 README.md                                     |    2 +
 atman-magma/atman_magma/openimages_eval.py    |    3 +-
 atman-magma/eval_openimages.py                |   96 -
 atman-magma/example_explain_panda_captum.py   |    3 +-
 atman-open-images-eval/.gitignore             |    8 -
 atman-open-images-eval/README.md              |   32 -
 atman-open-images-eval/config.yml             |   20 -
 .../multimodal_explain_eval/__init__.py       |    0
 .../multimodal_explain_eval/api.py            |  100 -
 .../multimodal_explain_eval/classes.py        |  352 ---
 .../multimodal_explain_eval/data_prep.py      |  162 -
 .../multimodal_explain_eval/dataloader.py     |  109 -
 .../multimodal_explain_eval/dataset.py        |  132 -
 .../multimodal_explain_eval/eval.py           |  212 --
 .../multimodal_explain_eval/utils.py          |  322 --
 atman-open-images-eval/prep_data.py           |   71 -
 atman-open-images-eval/requirements.in        |    6 -
 atman-open-images-eval/requirements.txt       |  350 --
 atman-open-images-eval/run.py                 |  100 -
 atman-open-images-eval/setup.py               |   31 -
 atman-open-images-eval/setup.sh               |    5 -
 determined/determined_eval_wrapper.py         |   32 -
 determined/metadata.json                      | 2802 -----------------
 determined/run_atman.yml                      |   48 -
 determined/run_atman_layers.yml               |   43 -
 determined/run_eval.py                        |  120 -
 startup-hook.sh                               |    1 -
 27 files changed, 4 insertions(+), 5158 deletions(-)
 delete mode 100755 atman-magma/eval_openimages.py
 delete mode 100755 atman-open-images-eval/.gitignore
 delete mode 100755 atman-open-images-eval/README.md
 delete mode 100755 atman-open-images-eval/config.yml
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/__init__.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/api.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/classes.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/data_prep.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/dataloader.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/dataset.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/eval.py
 delete mode 100755 atman-open-images-eval/multimodal_explain_eval/utils.py
 delete mode 100755 atman-open-images-eval/prep_data.py
 delete mode 100755 atman-open-images-eval/requirements.in
 delete mode 100755 atman-open-images-eval/requirements.txt
 delete mode 100755 atman-open-images-eval/run.py
 delete mode 100755 atman-open-images-eval/setup.py
 delete mode 100755 atman-open-images-eval/setup.sh
 delete mode 100755 determined/determined_eval_wrapper.py
 delete mode 100755 determined/metadata.json
 delete mode 100755 determined/run_atman.yml
 delete mode 100755 determined/run_atman_layers.yml
 delete mode 100755 determined/run_eval.py

diff --git a/README.md b/README.md
index 45d8105..7b2990c 100755
--- a/README.md
+++ b/README.md
@@ -24,6 +24,8 @@ bash startup-hook.sh
 ```
 Note: further model-checkpoints will be downloaded when executing for the first time. Sometimes CLIP fails to verify on the first execution -> running again works usually.
 
+The main folders are atman-magma, for all XAI implementations on the MAGMA model, and BLIP for all XAI implementations on the BLIP model.
+
 # examples with MAGMA
 ```
 cd atman-magma
diff --git a/atman-magma/atman_magma/openimages_eval.py b/atman-magma/atman_magma/openimages_eval.py
index 875c255..e04069d 100755
--- a/atman-magma/atman_magma/openimages_eval.py
+++ b/atman-magma/atman_magma/openimages_eval.py
@@ -3,7 +3,6 @@
 from PIL import Image
 from typing import Callable
 
-from multimodal_explain_eval.utils import check_if_a_or_an_and_get_prefix
 from magma.image_input import ImageInput
 from .utils import create_folder_if_does_not_exist
 
@@ -100,7 +99,7 @@ def run_eval(
 
             prompt = [
                 input_image,
-                text_prompt + check_if_a_or_an_and_get_prefix(word=data["label"]),
+                text_prompt + "a ",
             ]
 
             if use_lowercase_target == True:
diff --git a/atman-magma/eval_openimages.py b/atman-magma/eval_openimages.py
deleted file mode 100755
index 7e28f5d..0000000
--- a/atman-magma/eval_openimages.py
+++ /dev/null
@@ -1,96 +0,0 @@
-from atman_magma.magma  import Magma
-from atman_magma.explainer import Explainer
-from atman_magma.logit_parsing import get_delta_cross_entropies
-from atman_magma.openimages_eval import run_eval
-
-
-from multimodal_explain_eval.dataloader import DataLoader
-from multimodal_explain_eval.utils import load_json_as_dict
-
-import yaml
-
-with open("config.yml", "r") as stream:
-    config = yaml.safe_load(stream)
-
-conc_sup_values = config["hyperparams"]["conc_sup_values"]
-suppression_factor_values = config["hyperparams"]["suppression_factor_values"]
-
-
-metadata = load_json_as_dict(
-    filename = config['files']['metadata_filename']
-)
-
-keys_to_delete = []
-for x in metadata.keys():
-    if metadata[x]['count'] == 0:
-        keys_to_delete.append(x)
-        
-for key in keys_to_delete:
-    del metadata[key]
-
-
-dataloader = DataLoader(
-    metadata=metadata
-)
-
-num_images = 0
-for key in metadata:
-    num_images += metadata[key]["count"]
-    # print(f"{key:<17}:", metadata[key]["count"])
-
-print(f"total {num_images} pairs across {len(metadata)} classes")
-
-num_total_explanations = (
-    len(conc_sup_values) * len(suppression_factor_values) * num_images
-)
-print(f"Will run a total of: {num_total_explanations} explanations")
-
-print('output folder names:')
-folder_names = []
-for x in conc_sup_values:
-    for y in suppression_factor_values:
-        folder_name = f"{config['files']['output_dir']}/con_sup_thres_{round(x, 2) if x is not None else None}_suppression_factor_{round(y, 2) if y is not None else None}"
-        print(folder_name)
-        folder_names.append(
-            folder_name
-        )
-
-print('loading model...')
-model = Magma.from_checkpoint(
-    checkpoint_path = config['model']['checkpoint_path'],
-    device = config['model']['device']
-)
-
-folder_idx = 0
-
-for conceptual_suppression_threshold in conc_sup_values:
-    for suppression_factor in suppression_factor_values:
-
-        ex = Explainer(
-            model = model, 
-            device = config['model']['device'], 
-            tokenizer = model.tokenizer, 
-            conceptual_suppression_threshold = conceptual_suppression_threshold,
-            suppression_factor = suppression_factor
-        )
-
-        run_eval(
-            explainer = ex,
-            metadata = metadata,
-            dataloader = dataloader,
-            logit_parsing_fn=get_delta_cross_entropies,
-            output_folder = folder_names[folder_idx],
-            max_batch_size = config['model']['max_batch_size'],
-            text_prompt = config['hyperparams']['prompt_text'],
-            use_lowercase_target=True,
-            auto_decide_a_or_an=True,
-            progress=True,
-            square_outputs = False,
-            num_total_explanations=num_total_explanations,
-            prompt_explain_indices = [i for i in range(144)],
-            save_configs_only=config['hyperparams']['save_configs_only'],
-            save_configs_dir='./configs'
-        )
-        folder_idx += 1
-
-print('eval complete :)')
\ No newline at end of file
diff --git a/atman-magma/example_explain_panda_captum.py b/atman-magma/example_explain_panda_captum.py
index f883660..05cbc95 100755
--- a/atman-magma/example_explain_panda_captum.py
+++ b/atman-magma/example_explain_panda_captum.py
@@ -4,7 +4,6 @@
 from atman_magma.captum_helper import (
     CaptumMagma,
 )
-from multimodal_explain_eval.utils import check_if_a_or_an_and_get_prefix
 import numpy as np
 from atman_magma.magma  import Magma
 from magma.image_input import ImageInput
@@ -39,7 +38,7 @@
 att_combined = np.zeros((12,12))
 for i in range(len(label_tokens)):
 
-    text_prompt = f"This is a picture of {check_if_a_or_an_and_get_prefix(targets.lower())} "
+    text_prompt = f"This is a picture of a "
     if i >= 1:
         text_prompt += model.tokenizer.decode(label_tokens[:i])
 
diff --git a/atman-open-images-eval/.gitignore b/atman-open-images-eval/.gitignore
deleted file mode 100755
index b752e22..0000000
--- a/atman-open-images-eval/.gitignore
+++ /dev/null
@@ -1,8 +0,0 @@
-.ipynb_checkpoints/
-*.ipynb
-RESULTS/
-__pycache__/
-*.jpg
-.vscode/
-*.zip
-*.json
diff --git a/atman-open-images-eval/README.md b/atman-open-images-eval/README.md
deleted file mode 100755
index 73a775d..0000000
--- a/atman-open-images-eval/README.md
+++ /dev/null
@@ -1,32 +0,0 @@
-# atman-open-images-eval
-
-evaluating ATMAN's performance on the open-images-v6 segmentation dataset
-
-
-# installation
-
-1. Make sure torch, torchvision and cudatoolkit are installed and working
-
-2. Install the Aleph-Alpha Transformer codebase [as shown here](https://gitlab.aleph-alpha.de/research/transformer#install-as-package)
-
-2. install dependencides
-
-```
-pip install pip-tools ## optional
-pip install -r requirements.txt
-```
-
-# running with generator codebase
-
-By default, it uses this metadata: `metadata/all_classes_max_200_per_class.json`
-
-```
-python3 example_no_api.py
-```
-
-# add new requirements
-
-Add the new requirement to requirements.in and then run:
-```
-pip-compile requirements.in
-```
\ No newline at end of file
diff --git a/atman-open-images-eval/config.yml b/atman-open-images-eval/config.yml
deleted file mode 100755
index ae886ba..0000000
--- a/atman-open-images-eval/config.yml
+++ /dev/null
@@ -1,20 +0,0 @@
-logfile: "logs_30b.log"
-hyperparams:
-  conc_sup_values:
-    - 0.6
-  suppression_factor_values:
-    - 0.1
-  prompt_text: 'This is a picture of '
-
-files:
-  result_dir: '/temp_30b' ## remove the trailing "/"
-  raw_jsons_folder: "/temp_json_30b"
-  metadata_filename: 'metadata/all_classes_max_200_per_class.json'
-
-generator:
-  tokenizer_file: "alpha-001-128k.json"
-  checkpoint_dir: "/nfs/scratch/data_tmp/aa-alpha-001-128k-multilingual-data-alpha-1_30B_magma/global_step100000/"
-  pipe_parallel_size: 4
-  max_batch_size: 200 # edit this to be as big as possible
-  normalize : false
-  square_outputs : false
diff --git a/atman-open-images-eval/multimodal_explain_eval/__init__.py b/atman-open-images-eval/multimodal_explain_eval/__init__.py
deleted file mode 100755
index e69de29..0000000
diff --git a/atman-open-images-eval/multimodal_explain_eval/api.py b/atman-open-images-eval/multimodal_explain_eval/api.py
deleted file mode 100755
index 5aa9a0c..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/api.py
+++ /dev/null
@@ -1,100 +0,0 @@
-import json
-import requests
-import numpy as np
-import matplotlib.pyplot as plt
-
-def complete(
-    api_token,
-    prompt="Markus likes steak",
-    maximum_tokens=20,
-    model="luminous-extended",
-    url="https://test.api.aleph-alpha.com/complete",
-):
-
-    payload = json.dumps(
-        {"model": model, "prompt": prompt, "maximum_tokens": maximum_tokens}
-    )
-
-    headers = {
-        "Authorization": f"Bearer {api_token}",
-        "Content-Type": "application/json",
-    }
-
-    response = requests.post(url, headers=headers, data=payload)
-    return response.json()
-
-
-def explain(
-    api_token,
-    prompt="Markus likes steak",
-    target=" and fries",
-    directional=False,
-    suppression_factor=0.1,
-    conceptual_suppression_threshold=0.8,
-    model="luminous-extended",
-    url="https://api.aleph-alpha.com/explain",
-):
-
-    payload = json.dumps(
-        {
-            "model": model,
-            "prompt": prompt,
-            "target": target,
-            "directional": directional,
-            "suppression_factor": suppression_factor,
-            "conceptual_suppression_threshold": conceptual_suppression_threshold,
-        }
-    )
-
-    headers = {
-        "Authorization": f"Bearer {api_token}",
-        "Content-Type": "application/json",
-    }
-
-    response = requests.post(url, headers=headers, data=payload)
-    return response.json()
-
-
-def visualize_result_text(
-    result,
-    target_token_index=0,
-    prompt_start_index=None,
-    prompt_end_index=None,
-    width=20,
-    height=10,
-    topk=None,
-    fontsize=22,
-    save_as="output.jpg",
-    xticks_rotation=0,
-):
-
-    selected_result = result["result"][target_token_index]
-    explanations = selected_result["explanations"][prompt_start_index:prompt_end_index]
-    target_token = selected_result["target_token_str"]
-
-    assert (
-        len(explanations) != 0
-    ), "length of explanations is 0, you probably cropped it too much"
-
-    prompt_tokens = [
-        f'{i}:{explanations[i]["token_str"]}' for i in range(len(explanations))
-    ]
-
-    values = [explanations[i]["value"] for i in range(len(explanations))]
-
-    if topk is not None:
-        indices = np.argsort(np.array(values)).astype(int)[::-1]
-        values = np.take(values, indices)[:topk]
-        prompt_tokens = np.take(prompt_tokens, indices)[:topk]
-
-    fig = plt.figure(figsize=(width, height))
-    plt.title(f"target: {target_token}", fontsize=fontsize)
-    plt.bar(
-        prompt_tokens,
-        values,
-    )
-    plt.tick_params(axis="x", rotation=xticks_rotation)
-    plt.xticks(fontsize=fontsize)
-    plt.grid()
-
-    fig.savefig(save_as)
diff --git a/atman-open-images-eval/multimodal_explain_eval/classes.py b/atman-open-images-eval/multimodal_explain_eval/classes.py
deleted file mode 100755
index 23a1f6c..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/classes.py
+++ /dev/null
@@ -1,352 +0,0 @@
-openimages_v6_segmentation_classes = [
-    "Adhesive tape",
-    "Aircraft",
-    "Airplane",
-    "Alarm clock",
-    "Alpaca",
-    "Ambulance",
-    "Apple",
-    "Armadillo",
-    "Artichoke",
-    "Axe",
-    "Backpack",
-    "Bagel",
-    "Balance beam",
-    "Ball",
-    "Balloon",
-    "Banana",
-    "Band-aid",
-    "Barge",
-    "Barrel",
-    "Baseball bat",
-    "Baseball glove",
-    "Bat (Animal)",
-    "Beaker",
-    "Bear",
-    "Beer",
-    "Bell pepper",
-    "Belt",
-    "Bicycle wheel",
-    "Billiard table",
-    "Binoculars",
-    "Bird",
-    "Blue jay",
-    "Bomb",
-    "Book",
-    "Boot",
-    "Bottle",
-    "Bottle opener",
-    "Bowl",
-    "Box",
-    "Boy",
-    "Bread",
-    "Briefcase",
-    "Broccoli",
-    "Bronze sculpture",
-    "Brown bear",
-    "Bull",
-    "Burrito",
-    "Bus",
-    "Bust",
-    "Cabbage",
-    "Cake",
-    "Calculator",
-    "Camel",
-    "Camera",
-    "Canary",
-    "Candle",
-    "Canoe",
-    "Cantaloupe",
-    "Car",
-    "Carnivore",
-    "Carrot",
-    "Cat",
-    "Cattle",
-    "Cello",
-    "Cheese",
-    "Cheetah",
-    "Chest of drawers",
-    "Chicken",
-    "Chisel",
-    "Chopsticks",
-    "Christmas tree",
-    "Clock",
-    "Clothing",
-    "Cocktail",
-    "Cocktail shaker",
-    "Coffee",
-    "Coffee cup",
-    "Coin",
-    "Common fig",
-    "Common sunflower",
-    "Computer keyboard",
-    "Computer mouse",
-    "Cookie",
-    "Cooking spray",
-    "Corded phone",
-    "Couch",
-    "Cowboy hat",
-    "Cricket ball",
-    "Crocodile",
-    "Croissant",
-    "Cucumber",
-    "Dagger",
-    "Diaper",
-    "Dice",
-    "Digital clock",
-    "Dog",
-    "Dog bed",
-    "Dolphin",
-    "Door handle",
-    "Doughnut",
-    "Dress",
-    "Drill (Tool)",
-    "Drink",
-    "Drinking straw",
-    "Duck",
-    "Dumbbell",
-    "Eagle",
-    "Elephant",
-    "Envelope",
-    "Eraser",
-    "Facial tissue holder",
-    "Falcon",
-    "Fedora",
-    "Filing cabinet",
-    "Fire hydrant",
-    "Fish",
-    "Flag",
-    "Flashlight",
-    "Flower",
-    "Flowerpot",
-    "Flute",
-    "Flying disc",
-    "Food processor",
-    "Football",
-    "Fox",
-    "Frog",
-    "Frying pan",
-    "Garden Asparagus",
-    "Giraffe",
-    "Girl",
-    "Glove",
-    "Goat",
-    "Goldfish",
-    "Golf ball",
-    "Goose",
-    "Grape",
-    "Grapefruit",
-    "Guacamole",
-    "Guitar",
-    "Hair dryer",
-    "Hair spray",
-    "Hamburger",
-    "Hammer",
-    "Hamster",
-    "Hand dryer",
-    "Handbag",
-    "Handgun",
-    "Harbor seal",
-    "Harmonica",
-    "Harpsichord",
-    "Hat",
-    "Heater",
-    "Hedgehog",
-    "High heels",
-    "Hippopotamus",
-    "Horse",
-    "Hot dog",
-    "Human body",
-    "Human ear",
-    "Human mouth",
-    "Ipod",
-    "Jaguar (Animal)",
-    "Jeans",
-    "Jet ski",
-    "Jug",
-    "Juice",
-    "Kangaroo",
-    "Kettle",
-    "Kitchen knife",
-    "Kite",
-    "Knife",
-    "Koala",
-    "Ladle",
-    "Laptop",
-    "Lemon",
-    "Leopard",
-    "Light switch",
-    "Lighthouse",
-    "Limousine",
-    "Lion",
-    "Lipstick",
-    "Lizard",
-    "Loveseat",
-    "Luggage and bags",
-    "Lynx",
-    "Magpie",
-    "Man",
-    "Mango",
-    "Maracas",
-    "Measuring cup",
-    "Microwave oven",
-    "Milk",
-    "Miniskirt",
-    "Missile",
-    "Mixing bowl",
-    "Mobile phone",
-    "Monkey",
-    "Motorcycle",
-    "Mouse",
-    "Muffin",
-    "Mug",
-    "Mule",
-    "Mushroom",
-    "Nail (Construction)",
-    "Orange",
-    "Ostrich",
-    "Otter",
-    "Oven",
-    "Owl",
-    "Oyster",
-    "Pancake",
-    "Panda",
-    "Paper cutter",
-    "Paper towel",
-    "Parrot",
-    "Pastry",
-    "Peach",
-    "Pear",
-    "Pen",
-    "Pencil case",
-    "Penguin",
-    "Person",
-    "Piano",
-    "Picture frame",
-    "Pig",
-    "Pillow",
-    "Pitcher (Container)",
-    "Pizza",
-    "Pizza cutter",
-    "Plastic bag",
-    "Platter",
-    "Polar bear",
-    "Pomegranate",
-    "Popcorn",
-    "Potato",
-    "Power plugs and sockets",
-    "Pressure cooker",
-    "Pretzel",
-    "Printer",
-    "Pumpkin",
-    "Punching bag",
-    "Rabbit",
-    "Raccoon",
-    "Racket",
-    "Radish",
-    "Raven",
-    "Red panda",
-    "Remote control",
-    "Reptile",
-    "Rhinoceros",
-    "Rocket",
-    "Roller skates",
-    "Rose",
-    "Rugby ball",
-    "Ruler",
-    "Sandwich",
-    "Saucer",
-    "Saxophone",
-    "Scarf",
-    "Scissors",
-    "Screwdriver",
-    "Sculpture",
-    "Sea lion",
-    "Sea turtle",
-    "Seat belt",
-    "Segway",
-    "Shark",
-    "Sheep",
-    "Shirt",
-    "Shorts",
-    "Shower",
-    "Skateboard",
-    "Skirt",
-    "Skull",
-    "Skunk",
-    "Skyscraper",
-    "Slow cooker",
-    "Snake",
-    "Snowmobile",
-    "Soap dispenser",
-    "Sock",
-    "Sofa bed",
-    "Sombrero",
-    "Sparrow",
-    "Spatula",
-    "Spoon",
-    "Squash (Plant)",
-    "Squirrel",
-    "Stapler",
-    "Starfish",
-    "Stop sign",
-    "Strawberry",
-    "Studio couch",
-    "Submarine sandwich",
-    "Suit",
-    "Suitcase",
-    "Sun hat",
-    "Surfboard",
-    "Swan",
-    "Swim cap",
-    "Swimwear",
-    "Sword",
-    "Table tennis racket",
-    "Tablet computer",
-    "Tank",
-    "Tap",
-    "Tart",
-    "Taxi",
-    "Tea",
-    "Teapot",
-    "Teddy bear",
-    "Tennis ball",
-    "Tennis racket",
-    "Tie",
-    "Tiger",
-    "Toaster",
-    "Toilet",
-    "Toilet paper",
-    "Tomato",
-    "Torch",
-    "Tortoise",
-    "Towel",
-    "Toy",
-    "Traffic light",
-    "Traffic sign",
-    "Train",
-    "Trousers",
-    "Truck",
-    "Turkey",
-    "Turtle",
-    "Van",
-    "Vase",
-    "Vehicle registration plate",
-    "Volleyball (Ball)",
-    "Waffle",
-    "Washing machine",
-    "Waste container",
-    "Watch",
-    "Watermelon",
-    "Whale",
-    "Wheel",
-    "Whiteboard",
-    "Wine",
-    "Winter melon",
-    "Wok",
-    "Woman",
-    "Woodpecker",
-    "Wrench",
-    "Zebra",
-    "Zucchini",
-]
diff --git a/atman-open-images-eval/multimodal_explain_eval/data_prep.py b/atman-open-images-eval/multimodal_explain_eval/data_prep.py
deleted file mode 100755
index 86a2750..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/data_prep.py
+++ /dev/null
@@ -1,162 +0,0 @@
-from .dataset import SegmentationDataset
-import os
-from tqdm import tqdm
-import numpy as np 
-import cv2
-import shutil
-
-from .utils import check_if_a_or_an_and_get_prefix
-
-def get_filenames_in_a_folder(folder: str):
-    """
-    returns the list of paths to all the files in a given folder
-    """
-
-    files = os.listdir(folder)
-    num_files = len(files)
-    """
-    - class A
-        - images
-            - 1.jpg
-            - 2.jpg
-        - masks
-            - 1.jpg
-            - 2.jpg
-
-    - class B
-        - images
-            - 1.jpg
-            - 2.jpg
-        - masks
-            - 1.jpg
-            - 2.jpg
-    """
-
-    filenames_sorted = [
-        f"{folder}/{x+1}.jpg" for x in range(num_files)
-    ]
-    return filenames_sorted
-
-def prepare_openimages_dataset(
-    segmentation_dataset: SegmentationDataset,
-    class_names: list,
-    output_folder = './openimages_cleaned',
-    num_samples_per_class = 100,
-    start_clean = False,
-    max_width_by_height = 1.2,
-    min_width_by_height = 0.8,
-    min_dim = 200
-):
-
-    if os.path.exists(output_folder):
-        print(f'folder: {output_folder} already exists with folders:\n {get_filenames_in_a_folder(output_folder)}')
-        if start_clean is True:
-            print('starting clean, deleting existing folder...')
-            shutil.rmtree(output_folder)
-            os.mkdir(output_folder)
-    else:
-        print(f'making folder: {output_folder}')
-        os.mkdir(output_folder)
-    
-    metadata = {}
-
-    for name in class_names:
-        metadata[name] = {
-            'count': 0,
-            'prefix': check_if_a_or_an_and_get_prefix(word = name),
-            'folder': {
-                'images': None,
-                'masks': None
-            }
-        }
-
-    """
-    output_folder/
-        dog/
-            images/
-                1.jpg
-                2.jpg
-            masks/
-                1.jpg
-                2.jpg
-        cat/
-            images/
-                1.jpg
-                2.jpg
-            masks/
-                1.jpg
-                2.jpg
-    """
-
-    for item in tqdm(segmentation_dataset.dataset, desc = 'preparing data:'):
-        stuff = segmentation_dataset.postprocess(item)
-        label = stuff['label']
-
-        if label in list(metadata.keys()):
-
-            label_dir = output_folder + '/' + label
-            image_dir = label_dir + '/images'
-            mask_dir = label_dir + '/masks'
-
-            if os.path.exists(label_dir) is False:
-                print(f'making folder: {label_dir}')
-                os.mkdir(label_dir)
-
-            if os.path.exists(image_dir) is False:
-                print(f'making folder: {image_dir}')
-                os.mkdir(image_dir)
-
-            if os.path.exists(mask_dir) is False:
-                print(f'making folder: {mask_dir}')
-                os.mkdir(mask_dir)
-
-            ## I know this is redundant, but we'll run it only ones, so this worrks :)
-            metadata[label]['folder']['images'] = image_dir
-            metadata[label]['folder']['masks'] = mask_dir
-
-            if metadata[label]['count'] >= num_samples_per_class:
-                continue
-
-            image = stuff['image']
-            mask = stuff['mask']
-
-            width, height = stuff['width'], stuff['height']
-
-            ratio = width/height
-
-            minimum_dim_on_image = min([width, height])
-
-            if min_width_by_height < ratio < max_width_by_height and minimum_dim_on_image >= min_dim:
-                metadata[label]['count'] = metadata[label]['count'] + 1
-                image_path = image_dir + f"/{metadata[label]['count']}.jpg"
-                mask_path = mask_dir + f"/{metadata[label]['count']}.jpg"
-                # print(f'filename {item.filepath}, label dir: {label_dir}, image_path: {image_path}')
-
-                if os.path.exists(image_path) is False:
-                    image.save(image_path)
-                    # print(f'saving: {image_path}')
-                else:
-                    print(f'already exists: {image_path}')
-                cv2.imwrite(mask_path, (mask*255).astype(np.uint8))
-            else:
-                # print(f'rejecting: {label} because bad ratio: {ratio} or shape lower than min dim: {width, height}')
-                continue
-
-    for label, data in metadata.items():
-
-        try:
-            assert data['count'] == num_samples_per_class, f"Expected {num_samples_per_class} samples for {label} but got {data['count']}"
-        except AssertionError:
-            print('WARNING:', f"Expected {num_samples_per_class} samples for {label} but got {data['count']}")
-
-        try:
-            assert len(get_filenames_in_a_folder(data['folder']['images'])) == num_samples_per_class, f"Got {len(get_filenames_in_a_folder(data['folder']['images']))} images but expected {num_samples_per_class} images in folder: {get_filenames_in_a_folder(data['folder']['images'])}"
-        except AssertionError:
-            print('WARNING:', f"Got {len(get_filenames_in_a_folder(data['folder']['images']))} images but expected {num_samples_per_class} images in folder")
-        
-        try:
-            assert len(get_filenames_in_a_folder(data['folder']['masks'])) == num_samples_per_class, f"Got {len(get_filenames_in_a_folder(data['folder']['masks']))} masks but expected {num_samples_per_class} masks in folder: {get_filenames_in_a_folder(data['folder']['masks'])}"
-        except AssertionError:
-            print('WARNING:', f"Got {len(get_filenames_in_a_folder(data['folder']['masks']))} masks but expected {num_samples_per_class} masks in folder")
-
-    return metadata
\ No newline at end of file
diff --git a/atman-open-images-eval/multimodal_explain_eval/dataloader.py b/atman-open-images-eval/multimodal_explain_eval/dataloader.py
deleted file mode 100755
index a10d740..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/dataloader.py
+++ /dev/null
@@ -1,109 +0,0 @@
-from PIL import Image
-import cv2
-from .data_prep import get_filenames_in_a_folder
-import os 
-
-def center_crop_image(img, dim):
-	"""Returns center cropped image
-	Args:
-	img: image to be center cropped
-	dim: dimensions (width, height) to be cropped
-
-    source: https://gist.github.com/Nannigalaxy/35dd1d0722f29672e68b700bc5d44767
-	"""
-	width, height = img.shape[1], img.shape[0]
-
-	# process crop width and height for max available dimension
-	crop_width = dim[0] if dim[0]<img.shape[1] else img.shape[1]
-	crop_height = dim[1] if dim[1]<img.shape[0] else img.shape[0] 
-	mid_x, mid_y = int(width/2), int(height/2)
-	cw2, ch2 = int(crop_width/2), int(crop_height/2) 
-	crop_img = img[mid_y-ch2:mid_y+ch2, mid_x-cw2:mid_x+cw2]
-	return crop_img
-
-
-class DataLoader():
-    def __init__(self, metadata, load_image_on_getitem = False):
-        
-        """
-        metadata example:
-        metadata = {'Cat': {'count': 10,
-         'folder': {'images': './openimages_cleaned/Cat/images',
-                    'masks': './openimages_cleaned/Cat/masks'}},
-         'Dog': {'count': 10,
-                 'folder': {'images': './openimages_cleaned/Dog/images',
-                            'masks': './openimages_cleaned/Dog/masks'}}
-        }
-        """
-        self.load_image_on_getitem = load_image_on_getitem
-        self.metadata = metadata
-        self.labels = list(metadata.keys())
-        
-        self.image_paths = {}
-        self.mask_paths = {}
-        
-        for l in self.labels:
-            self.image_paths[l] = get_filenames_in_a_folder(self.metadata[l]['folder']['images'])
-            self.mask_paths[l] = get_filenames_in_a_folder(self.metadata[l]['folder']['masks'])
-
-        self.all_item_indices = self.prepare_stuff_for_iter(metadata=self.metadata)
-        print(f'dataloader contains a total of {len(self.all_item_indices)} indices')
-        
-    def fetch(self, label_idx = 0, idx = 0, center_crop = False, load_image = True):
-        assert label_idx < len(self.labels)
-        assert idx < len(self.image_paths[self.labels[label_idx]]), f'Got idx: {idx} but the highest was {len(self.image_paths[self.labels[label_idx]])}'
-        
-        image_path = self.image_paths[self.labels[label_idx]][idx]
-        assert os.path.exists(image_path), f'Expected image to exist in:{image_path}'
-        mask_path = self.mask_paths[self.labels[label_idx]][idx]
-    
-        if load_image == True:
-            image = cv2.cvtColor(cv2.imread(image_path), cv2.COLOR_BGR2RGB)
-        else:
-            image = None
-
-        mask = cv2.imread(mask_path)
-        if image is not None:
-            h, w, c = image.shape
-        else:
-            h, w, c = mask.shape
-
-        if center_crop is True:
-            min_dim = min([w,h])
-            if image is not None:
-                image  = center_crop_image(image, dim  = (min_dim, min_dim))
-            mask = center_crop_image(mask, dim  = (min_dim,min_dim))
-        
-        return {
-            'image': image,
-            'mask': mask,
-            'image_path': image_path,
-            "mask_path": mask_path,
-            'label': self.labels[label_idx]
-        }
-
-    def prepare_stuff_for_iter(self, metadata):
-        classes = list(metadata.keys())
-        all_item_indices = []
-
-        for i in range(len(classes)):
-            for j in range(metadata[classes[i]]["count"]):
-                all_item_indices.append(
-                    {
-                        'label_idx': i,
-                        'item_idx': j
-                    }
-                )
-
-        return all_item_indices
-
-    def __len__(self):
-        return len(self.all_item_indices)
-
-    def __getitem__(self, idx):
-        return self.fetch(
-            label_idx= self.all_item_indices[idx]['label_idx'],
-            idx = self.all_item_indices[idx]['item_idx'],
-            load_image=self.load_image_on_getitem
-        )
-        
\ No newline at end of file
diff --git a/atman-open-images-eval/multimodal_explain_eval/dataset.py b/atman-open-images-eval/multimodal_explain_eval/dataset.py
deleted file mode 100755
index f247655..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/dataset.py
+++ /dev/null
@@ -1,132 +0,0 @@
-import fiftyone.zoo as foz
-from PIL import Image
-import numpy as np
-import cv2
-
-class SegmentationDataset:
-    def __init__(self, 
-        classes, 
-        dataset_dir, 
-        num_total_samples = 2000,
-        width = 384, 
-        height = 384,
-        split = 'validation',
-        dataset_name = 'open-images-animals'
-    ):
-        self.classes= classes
-        self.num_total_samples = num_total_samples
-        self.dataset_dir = dataset_dir
-        self.width = width
-        self.height = height
-
-        self.dataset = foz.load_zoo_dataset(
-            "open-images-v6",
-            split=split,
-            label_types=["segmentations"],
-            classes = self.classes,
-            max_samples=self.num_total_samples,
-            seed=51,
-            shuffle=True,
-            dataset_name=dataset_name,
-            dataset_dir=self.dataset_dir
-        )     
-
-    def postprocess(self, item):
-        image = Image.open(item.filepath)
-        
-        masks = []
-
-        crop_data = {
-            'lefts': [],
-            'rights': [],
-            'tops': [],
-            'bottoms': []
-        }
-        
-
-        labels = [detection.label for detection in item.segmentations.detections]
-        indices_of_labels_same_as_first_one = []
-
-        for i in range(len(labels)):
-            if labels[i] == labels[0]:
-                indices_of_labels_same_as_first_one.append(i)
-
-        # print(f'found {len(indices_of_labels_same_as_first_one)} masks for {labels[0]}: {item.filepath}\n with all labels: {labels}')
-
-        for index in indices_of_labels_same_as_first_one:
-            detection = item.segmentations.detections[index]
-            # i think masks are relative, i.e. [0, 1.0] coordinates
-            box = detection.bounding_box
-            # not sure about the shape of the mask
-            mask = detection.mask
-            label = detection.label
-
-            left, right, top, bottom = get_crop_params(
-                image = image,
-                bounding_box = box
-            )
-
-            crop_data['lefts'].append(left)
-            crop_data['rights'].append(right)
-            crop_data['tops'].append(top)
-            crop_data['bottoms'].append(bottom)
-
-            cropped_image = image.crop((left, top, right, bottom))
-            
-            width, height = cropped_image.size
-            mask = mask.astype(np.float32)
-            mask = cv2.resize(mask, (width, height))
-
-            masks.append(mask)
-
-        final_left = min(crop_data['lefts'])
-        final_right = max(crop_data['rights'])
-
-        final_top = min(crop_data['tops'])
-        final_bottom = max(crop_data['bottoms'])
-
-        width, height = np.array(image).shape[0], np.array(image).shape[1]
-        output_mask = np.zeros((width, height)).astype(np.float32)
-
-
-        for i in range(len(masks)):
-            mask = masks[i]
-            left = crop_data['lefts'][i]
-            right = crop_data['rights'][i]
-            top = crop_data['tops'][i]
-            bottom = crop_data['bottoms'][i]
-            output_mask[top:bottom, left:right] += mask
-
-        cropped_image = image.crop((final_left, final_top, final_right, final_bottom))
-        cropped_output_mask = output_mask[final_top:final_bottom, final_left:final_right]
-
-        if self.height is not None and self.width is not None:
-            cropped_image = cropped_image.crop((left, top, right, bottom)).resize((self.height, self.width))
-            cropped_output_mask = cv2.resize(cropped_output_mask, dsize = (self.width, self.height))
-
-        width, height = cropped_image.size
-        return {
-                    'image': cropped_image,
-                    'mask': cropped_output_mask,
-                    'label': label,
-                    'filename': item.filepath,
-                    'width': width,
-                    'height': height
-                }
-        
-
-                
-
-def get_crop_params(image, bounding_box):
-    if np.array(image).ndim == 3:
-        height, width, channels = np.array(image).shape
-    else:
-         height, width = np.array(image).shape
-        
-    x0, y0, w, h = bounding_box
-    
-    left = int(width * x0)
-    top = int(height * y0)
-    right = left + int(w * width)
-    bottom = top + int(h * height)
-    return left, right, top, bottom
\ No newline at end of file
diff --git a/atman-open-images-eval/multimodal_explain_eval/eval.py b/atman-open-images-eval/multimodal_explain_eval/eval.py
deleted file mode 100755
index 7bd0c5c..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/eval.py
+++ /dev/null
@@ -1,212 +0,0 @@
-import os
-from PIL import Image
-from tqdm import tqdm
-import cv2
-import numpy as np
-import torch
-
-
-from .dataloader import DataLoader
-from .utils import (
-    result_to_segmentation_mask,
-    check_if_a_or_an_and_get_prefix,
-    dict_to_json,
-)
-
-from generator import TaskExplanation, TaskEvaluation, TaskCompletion
-
-
-def get_embedding(
-    generator_context,
-    text="hello world",
-):
-
-    with generator_context as generator:
-        token_ids = torch.tensor(generator.tokenizer.encode(text)).cuda()
-        emb = generator.transformer.transformer.embeddings.word_embeddings(token_ids)
-
-    return emb
-
-
-def get_logprob(
-    generator,
-    text_prompt,
-    pil_image,
-    label,
-    use_lowercase_target=True,
-    auto_decide_a_or_an=True,
-    also_collect_completion=True,
-):
-
-    if auto_decide_a_or_an == True:
-        if text_prompt[-1] != " ":
-            text_prompt += " "
-
-    prompt = [pil_image, text_prompt + check_if_a_or_an_and_get_prefix(word=label)]
-
-    if use_lowercase_target == True:
-        target = f" {label.lower()}"
-    else:
-        target = f" {label}"
-
-    # print(prompt, target)
-
-    task = TaskEvaluation(prompt=prompt, completion_expected=target)
-
-    output = generator.process([task])
-    data = {
-        "original_completion": None,
-        "logprob_on_target": output[0]["result"]["log_probability"],
-    }
-
-    if also_collect_completion is not None:
-        task = TaskCompletion(prompt=prompt, maximum_tokens=5)
-
-        output = generator.process([task])
-
-        data["original_completion"] = output[0]["completions"][0]["completion"]
-
-    return data
-
-
-def run_eval(
-    generator,
-    result_folder,
-    metadata,
-    dataloader: DataLoader,
-    text_prompt="This is a picture of ",  ## a or an is decided later
-    suppression_factor=0.1,
-    conceptual_suppression_threshold=0.5,
-    max_batch_size=32,
-    use_lowercase_target=True,
-    auto_decide_a_or_an=True,
-    progress=False,
-    raw_jsons_folder: str = None,
-    normalize: bool = False,
-    square_outputs: bool = False,
-    num_total_explanations=None,
-    logger = None
-):
-
-    if progress is True and num_total_explanations is not None:
-        pbar = tqdm(total=num_total_explanations)
-
-    """
-    ./result_folder
-        - Cat
-            - 1.jpg
-            - 2.jpg
-        - Dog
-            - 1.jpg
-            - 2.jpg
-    """
-
-    ## append a space between last word and a/an
-    if auto_decide_a_or_an == True:
-        if text_prompt[-1] != " ":
-            text_prompt += " "
-
-    RESULTS = {}
-
-    classes = list(metadata.keys())
-
-    d = dataloader
-
-    if os.path.exists(result_folder) is False:
-        print(f"making folder: {result_folder}")
-        os.mkdir(result_folder)
-
-    if os.path.exists(raw_jsons_folder) is False:
-        print(f"making folder: {raw_jsons_folder}")
-        os.mkdir(raw_jsons_folder)
-
-    for i in range(len(classes)):
-        label_dir = result_folder + "/" + classes[i]
-        json_dir = raw_jsons_folder + "/" + classes[i]
-
-        RESULTS[classes[i]] = {"folder": label_dir, "count": 0}
-
-        if os.path.exists(label_dir) is False:
-            print(f"making folder: {label_dir}")
-            os.mkdir(label_dir)
-
-        if os.path.exists(json_dir) is False and raw_jsons_folder is not None:
-            print(f"making folder: {json_dir}")
-            os.mkdir(json_dir)
-
-        for j in range(metadata[classes[i]]["count"]):
-
-            # 1 indexed filenames (starts with 1.jpg)
-            filename = label_dir + f"/{j+1}.jpg"
-            json_filename = json_dir + f"/{j+1}.json"
-
-            if os.path.exists(filename):
-                print(f"{filename} already exists, skipping...")
-                RESULTS[classes[i]]["count"] += 1
-                pbar.update(1)
-                continue
-
-            try:
-                data = d.fetch(i, j, center_crop=True)
-            except:
-                print(
-                    f"an error occured while fetching from dataloader: class: {classes[i]} idx: {j+1}"
-                )
-
-                if logger is not None:
-                    logger.info(
-                    f"an error occured while fetching from dataloader: class: {classes[i]} idx: {j+1}"
-                )
-                pbar.update(1)
-                continue
-
-            input_image = Image.fromarray(data["image"])
-
-            prompt = [
-                input_image,
-                text_prompt + check_if_a_or_an_and_get_prefix(word=data["label"]),
-            ]
-
-            if use_lowercase_target == True:
-                target = f" {data['label'].lower()}"
-            else:
-                target = f" {data['label']}"
-
-            task = TaskExplanation(
-                prompt=prompt,
-                target=target,
-                suppression_factor=suppression_factor,
-                conceptual_suppression_threshold=conceptual_suppression_threshold,
-                normalize=normalize,
-                square_outputs=square_outputs,
-            )
-
-            result = generator.process([task], max_batch_size=max_batch_size)[0]
-
-            mask = result_to_segmentation_mask(
-                result,
-                target_token_idx=[i for i in range(len(result["result"]))],
-                width=384,
-                height=384,
-            )
-
-            cv2.imwrite(filename, (mask * 255).astype(np.uint8))
-            if logger is not None:
-                logger.info(f'saved: {filename}')
-            else:
-                print(f"saved: {filename}")
-
-            if raw_jsons_folder is not None:
-                result["explanation_filename"] = filename
-
-                dict_to_json(dictionary=result, filename=json_filename)
-
-                if logger is not None:
-                    logger.info(f"saved json: {json_filename}")
-                else:
-                    print(f"saved json: {json_filename}")
-
-            RESULTS[classes[i]]["count"] += 1
-            pbar.update(1)
-    pbar.close()
-    return RESULTS
diff --git a/atman-open-images-eval/multimodal_explain_eval/utils.py b/atman-open-images-eval/multimodal_explain_eval/utils.py
deleted file mode 100755
index 08fde43..0000000
--- a/atman-open-images-eval/multimodal_explain_eval/utils.py
+++ /dev/null
@@ -1,322 +0,0 @@
-import matplotlib.pyplot as plt
-import cv2
-import numpy as np
-import json
-
-
-def load_json_as_dict(filename: str):
-    json1_file = open(filename)
-    json1_str = json1_file.read()
-    d = json.loads(json1_str)
-    return d
-
-
-def check_if_a_or_an_and_get_prefix(word: str):
-    vowels = ["a", "e", "i", "o", "u"]
-    if word[0].lower() in vowels:
-        return "an"
-    else:
-        return "a"
-
-
-def normalize(x, eps=1e-8):
-    """
-    values are shifted and rescaled so that they end up ranging between 0 and 1
-    """
-    normalized_array = (x - x.min()) / (x.max() - x.min()) + eps
-    return normalized_array
-
-
-def get_superimposed_heatmap(image, heatmap, colormap=cv2.COLORMAP_VIRIDIS):
-    colored_heatmap = cv2.applyColorMap(heatmap, colormap)
-    return 0.5 * image / 255.0 + 0.5 * (colored_heatmap / 255.0)
-
-
-def result_to_segmentation_mask(
-    result, target_token_idx: list, width=None, height=None
-):
-    values = []
-    dict_values = []
-    for idx in target_token_idx:
-
-        try:
-            dict_values.append(
-                result["result"][idx]["explanations"][:144]
-            )  ## make it to "explanations" for the API
-            values.append(
-                np.array([x["importance"] for x in dict_values[-1]]).reshape(12, 12)
-            )  ## value -> "importance" for the API
-        except KeyError:
-            dict_values.append(result["result"][idx]["explanation"][:144])
-            values.append(
-                np.array([x["value"] for x in dict_values[-1]]).reshape(12, 12)
-            )
-
-    for v in values:
-        v[0, 0] = v.mean()
-        v[-1, -1] = v.mean()
-
-    ## interpolation = cv2.INTER_NEAREST
-
-    values = sum(values) / len(values)
-
-    if width is not None and height is not None:
-        values = cv2.resize(values, (width, height), interpolation=cv2.INTER_NEAREST)
-    return values
-
-
-def visualize_result(
-    prompt_image_pil,
-    prompt,
-    target,
-    result,
-    target_token_idx,
-    suppression_factor,
-    conceptual_suppression_threshold,
-    fontsize=20,
-    colormap=cv2.COLORMAP_VIRIDIS,
-    sigma=None,
-):
-    dict_values = []
-    values = []
-    for idx in target_token_idx:
-        dict_values.append(result["result"][idx]["explanations"][:144])
-        values.append(
-            np.array([x["importance"] for x in dict_values[-1]]).reshape(12, 12)
-        )
-
-    for v in values:
-        v[0, 0] = v.mean()
-        v[-1, -1] = v.mean()
-
-    ## interpolation = cv2.INTER_NEAREST
-    values = cv2.resize(
-        sum(values) / len(values), (384, 384), interpolation=cv2.INTER_NEAREST
-    ).reshape(384, 384, 1)
-
-    if sigma is not None:
-        values = values**sigma
-
-    out = get_superimposed_heatmap(
-        image=cv2.resize(
-            np.array(prompt_image_pil),
-            dsize=(384, 384),
-        ),
-        heatmap=(normalize(values) * 255).astype(np.uint8),
-        colormap=colormap,
-    )
-
-    input_image = cv2.resize(
-        np.array(prompt_image_pil),
-        dsize=(384, 384),
-    )
-
-    fig, ax = plt.subplots(nrows=1, ncols=4, figsize=(30, 9))
-    fig.suptitle(f"PROMPT: <IMAGE> {prompt[-1]}\ntarget: {target}", fontsize=fontsize)
-
-    ax[0].set_title("input image")
-    ax[0].imshow(input_image)
-
-    ax[1].set_title(
-        f"target token: {''.join([result['result'][idx]['target_token_str'] for idx in target_token_idx])} \nsuppression_factor: {suppression_factor}\nconceptual_suppression_threshold: {conceptual_suppression_threshold}",
-        fontsize=fontsize,
-    )
-    ax[1].imshow(out)
-
-    im = ax[2].imshow(values)
-    ax[2].set_title(f"relative importance", fontsize=fontsize)
-
-    im = ax[3].imshow(values, vmin=0, vmax=1)
-    ax[3].set_title(
-        f"non normalized\n max:{round(values.max(), 5)}\n min:{round(values.min(), 5)}",
-        fontsize=fontsize,
-    )
-
-    fig.colorbar(im, ax=ax)
-
-    return input_image, values, out
-
-
-import base64
-from io import BytesIO
-
-
-def image_to_base64(image):
-    buffered = BytesIO()
-    image.save(buffered, format="JPEG")
-    buffer = buffered.getvalue()
-    return buffer
-
-
-def decode(jpg_as_text):
-    pil_image = Image.open(BytesIO(base64.b64decode(jpg_as_text)))
-    return pil_image
-
-
-def dict_to_json(dictionary, filename):
-    with open(filename, "w") as fp:
-        json.dump(dictionary, fp, indent=4)
-
-
-def calculate_time(
-    num_classes,
-    num_images_per_class,
-    num_conceptual_suppression_thresholds,
-    num_seconds_per_sample,
-):
-
-    t = round((
-        num_classes*num_images_per_class*num_conceptual_suppression_thresholds*num_seconds_per_sample
-    )/3600, 4)
-
-    print(
-        f'''\nnum_classes: {num_classes}
-num_images_per_class: {num_images_per_class}
-num_conceptual_suppression_thresholds: {num_conceptual_suppression_thresholds}
-num_seconds_per_sample: {num_seconds_per_sample}
-
-time estimate: {t} hours
----------------------------------------------------------------\n
-'''
-    )
-    return t 
-
-
-from typing import Callable
-
-def parse_results_for_single_image(
-    folder: str,
-    class_name: str,
-    idx: int,
-    dataloader,
-    metric_fn: Callable,
-    minimum_image_size: int = 384,
-    min_w_by_h: float = 0.9,
-    max_w_by_h: float = 1.1,
-    threshold: float = 0.1,
-    resize_to_12: bool = True,
-    result_dict: dict = None
-):
-    '''
-    result_dict looks like this:
-    SCORES_FOR_EACH_FILENAME = {
-        'image_filename': [],
-        'mask_filename': [],
-        'explanation_filename': [],
-        'precision_score': [],
-        'label': [],
-    }
-    '''
-    classes = list(dataloader.metadata.keys())
-
-    data = dataloader.fetch(classes.index(class_name), idx, center_crop=True, load_image = False)
-    y = data["mask"]
-    y = cv2.cvtColor(y, cv2.COLOR_BGR2GRAY)
-    y = y / y.max()
-
-    (
-        w,
-        h,
-    ) = y.shape
-
-    min_dim = min([w, h])
-
-    ratio = w / h
-
-    ## should have nice aspect ratio and big size, only then do we calculate precision score, else we return None
-    if min_dim > minimum_image_size and min_w_by_h < ratio < max_w_by_h:
-
-        ## its idx + 1 because the filenames were 1 indexed
-        explanation_path = folder + "/" + class_name + f"/{idx+1}.jpg"
-        x = cv2.imread(explanation_path)
-        x = cv2.cvtColor(x, cv2.COLOR_BGR2GRAY)
-        x = cv2.resize(x, dsize=(w, h))
-        x = x / (x.max() + 1e-8)
-
-        if threshold is not None:
-            x[x < threshold] = 0.0
-
-        ## resize to 12x12 grids
-        if resize_to_12 is True:
-            x = cv2.resize(x, (12, 12))
-            y = cv2.resize(y, (12, 12))
-        score = metric_fn(x, y)
-        
-        if result_dict is not None:
-            ## keeping scores for each image file
-            result_dict['image_filename'].append(
-                data['image_path']
-            )
-            result_dict['mask_filename'].append(
-                data['mask_path']
-            )
-            result_dict['explanation_filename'].append(
-                explanation_path
-            )
-            result_dict['precision_score'].append(
-                score
-            )
-            
-            result_dict['label'].append(
-                class_name
-            )   
-
-        return score
-    else:
-        return None
-
-from tqdm import tqdm
-import os 
-import pandas as pd
-
-def calculate_eval_scores_from_result_folder(
-    folder,
-    output_csv_file,
-    dataloader,
-    metric_fn,
-    minimum_image_size = 384,
-    min_w_by_h = 0.1,
-    max_w_by_h = 1.5,
-    threshold = 0.1,
-    resize_to_12 = True,
-):
-
-    result_dict = {
-        'image_filename': [],
-        'mask_filename': [],
-        'explanation_filename': [],
-        'precision_score': [],
-        'label': [],
-    }
-
-
-    for class_name in tqdm(list(dataloader.metadata.keys())):
-        num_images = dataloader.metadata[class_name]['count']
-        for idx in tqdm(range(num_images), desc = f'{class_name}', disable = True):
-            # calculate score
-            try:
-                ## making sure the explanation filename exists :)
-                filename = folder + "/" + class_name + f"/{idx+1}.jpg"
-                assert os.path.exists(filename), f'Image {filename} does not exist'
-
-                score = parse_results_for_single_image(
-                    result_dict = result_dict,
-                    folder = folder,
-                    class_name = class_name,
-                    idx = idx,
-                    dataloader =dataloader,
-                    metric_fn = metric_fn,
-                    minimum_image_size = minimum_image_size,
-                    min_w_by_h = min_w_by_h,
-                    max_w_by_h = max_w_by_h,
-                    threshold = threshold,
-                    resize_to_12 = resize_to_12
-                )
-            except:
-                print(f'An error occured for class_name: {class_name} idx: {idx}')
-
-    df = pd.DataFrame(result_dict)
-    df.to_csv(output_csv_file)
-    print(f'saved: {output_csv_file}')
-
-    return df
\ No newline at end of file
diff --git a/atman-open-images-eval/prep_data.py b/atman-open-images-eval/prep_data.py
deleted file mode 100755
index ef06d23..0000000
--- a/atman-open-images-eval/prep_data.py
+++ /dev/null
@@ -1,71 +0,0 @@
-from pprint import pprint
-
-from multimodal_explain_eval.data_prep import prepare_openimages_dataset
-from multimodal_explain_eval.dataset import SegmentationDataset
-
-from multimodal_explain_eval.classes import openimages_v6_segmentation_classes
-from pprint import pprint
-import json
-
-# CLASSES = [
-#     'Car',
-#     'Bus',
-#     'Cat',
-#     'Dog',
-#     'Pizza',
-#     'Hamburger',
-#     'Tiger',
-#     'Tortoise',
-#     'Chicken',
-#     'Calculator',
-#     'Clock',
-#     'Toilet',
-#     'Zebra'
-# ]
-
-CLASSES = openimages_v6_segmentation_classes
-
-
-print(f'Preparing: {len(CLASSES)} classes')
-
-ORIGINAL_DATASET_DIR = '/open-images-atman-eval'
-CLEAN_DATASET_DIR = '/openimages-cleaned-all-classes'
-
-all_metadata = {}
-
-for class_name in CLASSES:
-
-    dataset = SegmentationDataset(
-        classes = [class_name],
-        dataset_dir = ORIGINAL_DATASET_DIR,
-        num_total_samples = 1000, # change to a large number before final run
-        width = None,
-        height = None,
-        split = 'train',  ## change to train before final run
-        dataset_name = f'open-images-animals-{class_name}'
-    )
-
-    metadata = prepare_openimages_dataset(
-        segmentation_dataset = dataset,
-        class_names = [class_name],
-        output_folder = CLEAN_DATASET_DIR,
-        num_samples_per_class = 200,
-        start_clean = False,
-        max_width_by_height = 1.2,
-        min_width_by_height = 0.8,
-        min_dim = 200
-    )
-    key = list(metadata.keys())[0]
-    all_metadata[key] = metadata[key]
-
-pprint(all_metadata)
-
-with open('metadata.json', 'w') as fp:
-    json.dump(all_metadata, fp, indent = 4)
-
-count = 0
-
-for key in all_metadata:
-    count += all_metadata[key]['count']
-
-print(f'Saved a total of: {count} images in {CLEAN_DATASET_DIR}')
diff --git a/atman-open-images-eval/requirements.in b/atman-open-images-eval/requirements.in
deleted file mode 100755
index 71c96ec..0000000
--- a/atman-open-images-eval/requirements.in
+++ /dev/null
@@ -1,6 +0,0 @@
-pip-tools
-aleph-alpha-client
-datasets
-opencv-python-headless
-pillow
-fiftyone
\ No newline at end of file
diff --git a/atman-open-images-eval/requirements.txt b/atman-open-images-eval/requirements.txt
deleted file mode 100755
index af9f2bb..0000000
--- a/atman-open-images-eval/requirements.txt
+++ /dev/null
@@ -1,350 +0,0 @@
-#
-# This file is autogenerated by pip-compile with python 3.9
-# To update, run:
-#
-#    pip-compile requirements.in
-#
-aiofiles==0.8.0
-    # via fiftyone
-aiohttp==3.8.1
-    # via
-    #   datasets
-    #   fsspec
-aiosignal==1.2.0
-    # via aiohttp
-aleph-alpha-client==1.5.0
-    # via -r requirements.in
-anyio==3.6.1
-    # via
-    #   httpcore
-    #   starlette
-argcomplete==2.0.0
-    # via
-    #   fiftyone
-    #   voxel51-eta
-async-timeout==4.0.2
-    # via aiohttp
-attrs==21.4.0
-    # via aiohttp
-backports-cached-property==1.0.2
-    # via strawberry-graphql
-boto3==1.24.21
-    # via fiftyone
-botocore==1.27.21
-    # via
-    #   boto3
-    #   s3transfer
-certifi==2022.6.15
-    # via
-    #   httpcore
-    #   httpx
-    #   requests
-charset-normalizer==2.0.12
-    # via
-    #   aiohttp
-    #   requests
-click==8.1.3
-    # via
-    #   pip-tools
-    #   strawberry-graphql
-cycler==0.11.0
-    # via matplotlib
-dacite==1.6.0
-    # via fiftyone
-datasets==2.3.2
-    # via -r requirements.in
-deprecated==1.2.13
-    # via fiftyone
-dill==0.3.5.1
-    # via
-    #   datasets
-    #   multiprocess
-    #   voxel51-eta
-dnspython==2.2.1
-    # via eventlet
-eventlet==0.33.1
-    # via fiftyone
-fiftyone==0.16.5
-    # via -r requirements.in
-fiftyone-brain==0.8.2
-    # via fiftyone
-fiftyone-db==0.3.0
-    # via fiftyone
-filelock==3.7.1
-    # via huggingface-hub
-fonttools==4.33.3
-    # via matplotlib
-frozenlist==1.3.0
-    # via
-    #   aiohttp
-    #   aiosignal
-fsspec[http]==2022.5.0
-    # via datasets
-future==0.18.2
-    # via
-    #   fiftyone
-    #   voxel51-eta
-glob2==0.7
-    # via voxel51-eta
-graphql-core==3.1.7
-    # via strawberry-graphql
-greenlet==1.1.2
-    # via eventlet
-h11==0.12.0
-    # via
-    #   httpcore
-    #   hypercorn
-    #   wsproto
-h2==4.1.0
-    # via hypercorn
-hpack==4.0.0
-    # via h2
-httpcore==0.15.0
-    # via httpx
-httpx==0.23.0
-    # via universal-analytics-python3
-huggingface-hub==0.8.1
-    # via datasets
-hypercorn==0.13.2
-    # via fiftyone
-hyperframe==6.0.1
-    # via h2
-idna==3.3
-    # via
-    #   anyio
-    #   requests
-    #   rfc3986
-    #   yarl
-imageio==2.19.3
-    # via scikit-image
-jinja2==3.1.2
-    # via fiftyone
-jmespath==1.0.1
-    # via
-    #   boto3
-    #   botocore
-joblib==1.1.0
-    # via scikit-learn
-kaleido==0.2.1
-    # via fiftyone
-kiwisolver==1.4.3
-    # via matplotlib
-markupsafe==2.1.1
-    # via jinja2
-matplotlib==3.5.2
-    # via fiftyone
-mongoengine==0.20.0
-    # via fiftyone
-motor==2.5.1
-    # via fiftyone
-multidict==6.0.2
-    # via
-    #   aiohttp
-    #   yarl
-multiprocess==0.70.13
-    # via datasets
-ndjson==0.3.1
-    # via
-    #   fiftyone
-    #   voxel51-eta
-networkx==2.8.4
-    # via scikit-image
-numpy==1.22.4
-    # via
-    #   datasets
-    #   fiftyone
-    #   fiftyone-brain
-    #   imageio
-    #   matplotlib
-    #   opencv-python-headless
-    #   pandas
-    #   pyarrow
-    #   pywavelets
-    #   scikit-image
-    #   scikit-learn
-    #   scipy
-    #   tifffile
-    #   voxel51-eta
-opencv-python-headless==4.6.0.66
-    # via
-    #   -r requirements.in
-    #   fiftyone
-    #   voxel51-eta
-packaging==21.3
-    # via
-    #   datasets
-    #   fiftyone
-    #   huggingface-hub
-    #   matplotlib
-    #   scikit-image
-    #   voxel51-eta
-pandas==1.4.2
-    # via
-    #   datasets
-    #   fiftyone
-patool==1.12
-    # via voxel51-eta
-pep517==0.12.0
-    # via pip-tools
-pillow==9.1.1
-    # via
-    #   -r requirements.in
-    #   fiftyone
-    #   imageio
-    #   matplotlib
-    #   scikit-image
-    #   voxel51-eta
-pip-tools==6.6.2
-    # via -r requirements.in
-plotly==4.14.3
-    # via fiftyone
-pprintpp==0.4.0
-    # via fiftyone
-priority==2.0.0
-    # via hypercorn
-psutil==5.9.1
-    # via fiftyone
-pyarrow==8.0.0
-    # via datasets
-pygments==2.12.0
-    # via strawberry-graphql
-pymongo==3.12.3
-    # via
-    #   fiftyone
-    #   mongoengine
-    #   motor
-pyparsing==3.0.9
-    # via
-    #   matplotlib
-    #   packaging
-python-dateutil==2.8.2
-    # via
-    #   botocore
-    #   matplotlib
-    #   pandas
-    #   strawberry-graphql
-    #   voxel51-eta
-python-multipart==0.0.5
-    # via strawberry-graphql
-pytz==2022.1
-    # via
-    #   fiftyone
-    #   pandas
-    #   voxel51-eta
-pytz-deprecation-shim==0.1.0.post0
-    # via tzlocal
-pywavelets==1.3.0
-    # via scikit-image
-pyyaml==6.0
-    # via
-    #   fiftyone
-    #   huggingface-hub
-requests==2.28.0
-    # via
-    #   aleph-alpha-client
-    #   datasets
-    #   fsspec
-    #   huggingface-hub
-    #   responses
-    #   voxel51-eta
-responses==0.18.0
-    # via datasets
-retrying==1.3.3
-    # via
-    #   fiftyone
-    #   plotly
-    #   voxel51-eta
-rfc3986[idna2008]==1.5.0
-    # via httpx
-s3transfer==0.6.0
-    # via boto3
-scikit-image==0.19.3
-    # via
-    #   fiftyone
-    #   voxel51-eta
-scikit-learn==1.1.1
-    # via
-    #   fiftyone
-    #   fiftyone-brain
-scipy==1.8.1
-    # via
-    #   fiftyone-brain
-    #   scikit-image
-    #   scikit-learn
-sentinel==0.3.0
-    # via strawberry-graphql
-six==1.16.0
-    # via
-    #   eventlet
-    #   plotly
-    #   python-dateutil
-    #   python-multipart
-    #   retrying
-    #   voxel51-eta
-sniffio==1.2.0
-    # via
-    #   anyio
-    #   httpcore
-    #   httpx
-sortedcontainers==2.4.0
-    # via voxel51-eta
-sse-starlette==0.10.3
-    # via fiftyone
-sseclient-py==1.7.2
-    # via fiftyone
-starlette==0.16.0
-    # via
-    #   fiftyone
-    #   sse-starlette
-strawberry-graphql==0.96.0
-    # via fiftyone
-tabulate==0.8.10
-    # via
-    #   fiftyone
-    #   voxel51-eta
-threadpoolctl==3.1.0
-    # via scikit-learn
-tifffile==2022.5.4
-    # via scikit-image
-toml==0.10.2
-    # via hypercorn
-tomli==2.0.1
-    # via pep517
-tqdm==4.64.0
-    # via
-    #   datasets
-    #   huggingface-hub
-typing-extensions==4.2.0
-    # via
-    #   huggingface-hub
-    #   strawberry-graphql
-tzdata==2022.1
-    # via pytz-deprecation-shim
-tzlocal==4.2
-    # via voxel51-eta
-universal-analytics-python3==1.1.1
-    # via fiftyone
-urllib3==1.26.9
-    # via
-    #   botocore
-    #   requests
-    #   responses
-    #   voxel51-eta
-voxel51-eta==0.7.1
-    # via fiftyone
-wheel==0.37.1
-    # via pip-tools
-wrapt==1.14.1
-    # via deprecated
-wsproto==1.1.0
-    # via hypercorn
-xmltodict==0.13.0
-    # via fiftyone
-xxhash==3.0.0
-    # via datasets
-yarl==1.7.2
-    # via aiohttp
-
-# The following packages are considered to be unsafe in a requirements file:
-# pip
-# setuptools
diff --git a/atman-open-images-eval/run.py b/atman-open-images-eval/run.py
deleted file mode 100755
index 8a91694..0000000
--- a/atman-open-images-eval/run.py
+++ /dev/null
@@ -1,100 +0,0 @@
-from multimodal_explain_eval.eval import run_eval
-from multimodal_explain_eval.dataloader import DataLoader
-from multimodal_explain_eval.utils import load_json_as_dict
-
-import os
-import yaml
-import logging
-from generator import Generator
-
-with open("config.yml", "r") as stream:
-    config = yaml.safe_load(stream)
-
-logger = logging.getLogger()
-handler = logging.FileHandler(config['logfile'])
-logger.addHandler(handler)
-logger = logging.getLogger()
-logger.info(f"Running with config: {config}")
-
-
-conc_sup_values = config["hyperparams"]["conc_sup_values"]
-suppression_factor_values = config["hyperparams"]["suppression_factor_values"]
-RESULT_DIR = config["files"]["result_dir"]
-JSON_DIR = config["files"]["raw_jsons_folder"]
-metadata = load_json_as_dict(config["files"]["metadata_filename"])
-
-num_images = 0
-
-for key in metadata:
-    num_images += metadata[key]["count"]
-    # print(f"{key:<17}:", metadata[key]["count"])
-
-print(f"total {num_images} pairs across {len(metadata)} classes")
-
-num_total_explanations = (
-    len(conc_sup_values) * len(suppression_factor_values) * num_images
-)
-print(f"Will run a total of: {num_total_explanations} explanations")
-
-folder_names = []
-json_folder_names = []
-assert os.path.exists(RESULT_DIR) == True, f"Expected folder to exist:{RESULT_DIR}"
-
-for x in conc_sup_values:
-    for y in suppression_factor_values:
-        folder_names.append(
-            f"{RESULT_DIR}/con_sup_thres_{round(x, 2) if x is not None else None}_suppression_factor_{round(y, 2) if x is not None else None}"
-        )
-        json_folder_names.append(
-            f"{JSON_DIR}/con_sup_thres_{round(x, 2) if x is not None else None}_suppression_factor_{round(y, 2) if x is not None else None}"
-        )
-
-print('FOLDER NAMES:')
-for f in folder_names:
-    print(f)
-
-print('-'*100)
-print('JSON FOLDER NAMES')
-for f in json_folder_names:
-    print(f)
-
-print("preparing dataloader...")
-dataloader = DataLoader(metadata=metadata)
-
-print("loading generator")
-generator_context = Generator.from_checkpoint(
-    tokenizer_file=config["generator"]["tokenizer_file"],
-    checkpoint_dir=config["generator"]["checkpoint_dir"],
-    pipe_parallel_size=config["generator"]["pipe_parallel_size"],
-)
-
-folder_idx = 0
-
-print("STARTED RUN, GO TO SLEEP zzz")
-
-with generator_context as generator:
-    for conceptual_suppression_threshold in conc_sup_values:
-        for suppression_factor in suppression_factor_values:
-
-            folder_name = folder_names[folder_idx]
-            result_metadata = run_eval(
-                generator=generator,
-                result_folder=folder_name,
-                logger = logger,
-                text_prompt=config["hyperparams"]["prompt_text"],
-                max_batch_size=config["generator"]["max_batch_size"],
-                suppression_factor=suppression_factor,
-                conceptual_suppression_threshold=conceptual_suppression_threshold,
-                dataloader=dataloader,
-                metadata=metadata,
-                use_lowercase_target=True,
-                auto_decide_a_or_an=True,
-                progress=True,
-                raw_jsons_folder=json_folder_names[folder_idx],
-                normalize=config["generator"]["normalize"],
-                square_outputs=config["generator"]["square_outputs"],
-		num_total_explanations=num_total_explanations
-            )
-            folder_idx += 1
-
-print("RUN COMPLETE, hope you slept well :)")
diff --git a/atman-open-images-eval/setup.py b/atman-open-images-eval/setup.py
deleted file mode 100755
index 2efa87b..0000000
--- a/atman-open-images-eval/setup.py
+++ /dev/null
@@ -1,31 +0,0 @@
-import setuptools
-
-with open("README.md", "r") as fh:
-    long_description = fh.read()
-
-with open('requirements.txt') as f:
-    required = f.read().splitlines()
-
-setuptools.setup(
-    name="atman-open-images-eval",
-    version="0.0.1",
-    author="Aleph-Alpha",
-    author_email="",
-    description= "open images dataset eval code for explainability benchmark",
-    long_description=long_description,
-    long_description_content_type="text/markdown",
-    url="",
-    packages=setuptools.find_packages(),
-    install_requires= required,
-    python_requires='>=3.6',
-    include_package_data=True,
-    keywords=[
-        "PyTorch",
-        "machine learning",
-        ],
-    classifiers=[
-        "Intended Audience :: Science/Research",
-    ],
-    test_suite='nose.collector',
-    tests_require=['nose']
-)
diff --git a/atman-open-images-eval/setup.sh b/atman-open-images-eval/setup.sh
deleted file mode 100755
index 10d1224..0000000
--- a/atman-open-images-eval/setup.sh
+++ /dev/null
@@ -1,5 +0,0 @@
-pip install -r requirements.txt
-pip install "opencv-python-headless<4.3"
-pip install gdown
-## download dataset: https://drive.google.com/file/d/1w-IWWnW8DzRrudZtojPibBP-LYvG4zxU/view?usp=sharing
-gdown 1w-IWWnW8DzRrudZtojPibBP-LYvG4zxU
diff --git a/determined/determined_eval_wrapper.py b/determined/determined_eval_wrapper.py
deleted file mode 100755
index 4364d82..0000000
--- a/determined/determined_eval_wrapper.py
+++ /dev/null
@@ -1,32 +0,0 @@
-from dataclasses import dataclass, fields
-
-
-@dataclass
-class EvalParams:
-    sup_fact: str = ".3"
-    conc_sup: str = ".3"
-    result_path: str = "/det_cos_nolog_mult"
-    layers: str=None
-
-
-def process_determined_hparams(hparams):
-    """
-    Process the 'hyperparameters:' block in the determined config. Returns a custom dataclass with the specified parameter values. Hyperparameters are optional; if provided they override the command-line args in the 'entrypoint' command; if not provided, they can also be set in 'entrypoint:' if common for all tasks/models or the default values in main.py are used.
-    """
-
-    param_mismatches = []
-
-    fieldSet = {
-        f.name if f.init else param_mismatches.append(f.name)
-        for f in fields(EvalParams)
-    }
-    if param_mismatches != []:
-        raise KeyError(
-            f"The Parameters: {param_mismatches} are not included in the Dataclass."
-        )
-
-    filteredArgDict = {k: v for k, v in hparams.items() if k in fieldSet}
-    print("Filtered Arg Dict", filteredArgDict)
-    merged_params = EvalParams(**filteredArgDict)
-
-    return merged_params
diff --git a/determined/metadata.json b/determined/metadata.json
deleted file mode 100755
index 5c37129..0000000
--- a/determined/metadata.json
+++ /dev/null
@@ -1,2802 +0,0 @@
-{
-    "Adhesive tape": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Adhesive tape/images",
-            "masks": "/openimages-mini/Adhesive tape/masks"
-        }
-    },
-    "Aircraft": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Aircraft/images",
-            "masks": "/openimages-mini/Aircraft/masks"
-        }
-    },
-    "Airplane": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Airplane/images",
-            "masks": "/openimages-mini/Airplane/masks"
-        }
-    },
-    "Alarm clock": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Alarm clock/images",
-            "masks": "/openimages-mini/Alarm clock/masks"
-        }
-    },
-    "Alpaca": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Alpaca/images",
-            "masks": "/openimages-mini/Alpaca/masks"
-        }
-    },
-    "Ambulance": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Ambulance/images",
-            "masks": "/openimages-mini/Ambulance/masks"
-        }
-    },
-    "Apple": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Apple/images",
-            "masks": "/openimages-mini/Apple/masks"
-        }
-    },
-    "Armadillo": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Artichoke": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Artichoke/images",
-            "masks": "/openimages-mini/Artichoke/masks"
-        }
-    },
-    "Axe": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Backpack": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Backpack/images",
-            "masks": "/openimages-mini/Backpack/masks"
-        }
-    },
-    "Bagel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bagel/images",
-            "masks": "/openimages-mini/Bagel/masks"
-        }
-    },
-    "Balance beam": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Balance beam/images",
-            "masks": "/openimages-mini/Balance beam/masks"
-        }
-    },
-    "Ball": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Ball/images",
-            "masks": "/openimages-mini/Ball/masks"
-        }
-    },
-    "Balloon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Balloon/images",
-            "masks": "/openimages-mini/Balloon/masks"
-        }
-    },
-    "Banana": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Banana/images",
-            "masks": "/openimages-mini/Banana/masks"
-        }
-    },
-    "Band-aid": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Barge": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Barge/images",
-            "masks": "/openimages-mini/Barge/masks"
-        }
-    },
-    "Barrel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Barrel/images",
-            "masks": "/openimages-mini/Barrel/masks"
-        }
-    },
-    "Baseball bat": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Baseball bat/images",
-            "masks": "/openimages-mini/Baseball bat/masks"
-        }
-    },
-    "Baseball glove": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Baseball glove/images",
-            "masks": "/openimages-mini/Baseball glove/masks"
-        }
-    },
-    "Bat (Animal)": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bat (Animal)/images",
-            "masks": "/openimages-mini/Bat (Animal)/masks"
-        }
-    },
-    "Beaker": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Bear": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bear/images",
-            "masks": "/openimages-mini/Bear/masks"
-        }
-    },
-    "Beer": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Beer/images",
-            "masks": "/openimages-mini/Beer/masks"
-        }
-    },
-    "Bell pepper": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bell pepper/images",
-            "masks": "/openimages-mini/Bell pepper/masks"
-        }
-    },
-    "Belt": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Belt/images",
-            "masks": "/openimages-mini/Belt/masks"
-        }
-    },
-    "Bicycle wheel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bicycle wheel/images",
-            "masks": "/openimages-mini/Bicycle wheel/masks"
-        }
-    },
-    "Billiard table": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Billiard table/images",
-            "masks": "/openimages-mini/Billiard table/masks"
-        }
-    },
-    "Binoculars": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Binoculars/images",
-            "masks": "/openimages-mini/Binoculars/masks"
-        }
-    },
-    "Bird": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bird/images",
-            "masks": "/openimages-mini/Bird/masks"
-        }
-    },
-    "Blue jay": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Blue jay/images",
-            "masks": "/openimages-mini/Blue jay/masks"
-        }
-    },
-    "Bomb": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Book": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Book/images",
-            "masks": "/openimages-mini/Book/masks"
-        }
-    },
-    "Boot": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Boot/images",
-            "masks": "/openimages-mini/Boot/masks"
-        }
-    },
-    "Bottle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bottle/images",
-            "masks": "/openimages-mini/Bottle/masks"
-        }
-    },
-    "Bottle opener": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Bowl": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bowl/images",
-            "masks": "/openimages-mini/Bowl/masks"
-        }
-    },
-    "Box": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Box/images",
-            "masks": "/openimages-mini/Box/masks"
-        }
-    },
-    "Boy": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Boy/images",
-            "masks": "/openimages-mini/Boy/masks"
-        }
-    },
-    "Bread": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bread/images",
-            "masks": "/openimages-mini/Bread/masks"
-        }
-    },
-    "Briefcase": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Briefcase/images",
-            "masks": "/openimages-mini/Briefcase/masks"
-        }
-    },
-    "Broccoli": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Broccoli/images",
-            "masks": "/openimages-mini/Broccoli/masks"
-        }
-    },
-    "Bronze sculpture": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bronze sculpture/images",
-            "masks": "/openimages-mini/Bronze sculpture/masks"
-        }
-    },
-    "Brown bear": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Brown bear/images",
-            "masks": "/openimages-mini/Brown bear/masks"
-        }
-    },
-    "Bull": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bull/images",
-            "masks": "/openimages-mini/Bull/masks"
-        }
-    },
-    "Burrito": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Burrito/images",
-            "masks": "/openimages-mini/Burrito/masks"
-        }
-    },
-    "Bus": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bus/images",
-            "masks": "/openimages-mini/Bus/masks"
-        }
-    },
-    "Bust": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Bust/images",
-            "masks": "/openimages-mini/Bust/masks"
-        }
-    },
-    "Cabbage": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cabbage/images",
-            "masks": "/openimages-mini/Cabbage/masks"
-        }
-    },
-    "Cake": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cake/images",
-            "masks": "/openimages-mini/Cake/masks"
-        }
-    },
-    "Calculator": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Calculator/images",
-            "masks": "/openimages-mini/Calculator/masks"
-        }
-    },
-    "Camel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Camel/images",
-            "masks": "/openimages-mini/Camel/masks"
-        }
-    },
-    "Camera": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Camera/images",
-            "masks": "/openimages-mini/Camera/masks"
-        }
-    },
-    "Canary": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Canary/images",
-            "masks": "/openimages-mini/Canary/masks"
-        }
-    },
-    "Candle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Candle/images",
-            "masks": "/openimages-mini/Candle/masks"
-        }
-    },
-    "Canoe": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Canoe/images",
-            "masks": "/openimages-mini/Canoe/masks"
-        }
-    },
-    "Cantaloupe": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cantaloupe/images",
-            "masks": "/openimages-mini/Cantaloupe/masks"
-        }
-    },
-    "Car": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Car/images",
-            "masks": "/openimages-mini/Car/masks"
-        }
-    },
-    "Carnivore": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Carnivore/images",
-            "masks": "/openimages-mini/Carnivore/masks"
-        }
-    },
-    "Carrot": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Carrot/images",
-            "masks": "/openimages-mini/Carrot/masks"
-        }
-    },
-    "Cat": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cat/images",
-            "masks": "/openimages-mini/Cat/masks"
-        }
-    },
-    "Cattle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cattle/images",
-            "masks": "/openimages-mini/Cattle/masks"
-        }
-    },
-    "Cello": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cello/images",
-            "masks": "/openimages-mini/Cello/masks"
-        }
-    },
-    "Cheese": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cheese/images",
-            "masks": "/openimages-mini/Cheese/masks"
-        }
-    },
-    "Cheetah": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cheetah/images",
-            "masks": "/openimages-mini/Cheetah/masks"
-        }
-    },
-    "Chest of drawers": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Chest of drawers/images",
-            "masks": "/openimages-mini/Chest of drawers/masks"
-        }
-    },
-    "Chicken": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Chicken/images",
-            "masks": "/openimages-mini/Chicken/masks"
-        }
-    },
-    "Chisel": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Chisel/images",
-            "masks": "/openimages-mini/Chisel/masks"
-        }
-    },
-    "Chopsticks": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Chopsticks/images",
-            "masks": "/openimages-mini/Chopsticks/masks"
-        }
-    },
-    "Christmas tree": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Christmas tree/images",
-            "masks": "/openimages-mini/Christmas tree/masks"
-        }
-    },
-    "Clock": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Clock/images",
-            "masks": "/openimages-mini/Clock/masks"
-        }
-    },
-    "Clothing": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Clothing/images",
-            "masks": "/openimages-mini/Clothing/masks"
-        }
-    },
-    "Cocktail": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cocktail/images",
-            "masks": "/openimages-mini/Cocktail/masks"
-        }
-    },
-    "Cocktail shaker": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Coffee": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Coffee/images",
-            "masks": "/openimages-mini/Coffee/masks"
-        }
-    },
-    "Coffee cup": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Coffee cup/images",
-            "masks": "/openimages-mini/Coffee cup/masks"
-        }
-    },
-    "Coin": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Coin/images",
-            "masks": "/openimages-mini/Coin/masks"
-        }
-    },
-    "Common fig": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Common fig/images",
-            "masks": "/openimages-mini/Common fig/masks"
-        }
-    },
-    "Common sunflower": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Common sunflower/images",
-            "masks": "/openimages-mini/Common sunflower/masks"
-        }
-    },
-    "Computer keyboard": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Computer keyboard/images",
-            "masks": "/openimages-mini/Computer keyboard/masks"
-        }
-    },
-    "Computer mouse": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Computer mouse/images",
-            "masks": "/openimages-mini/Computer mouse/masks"
-        }
-    },
-    "Cookie": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cookie/images",
-            "masks": "/openimages-mini/Cookie/masks"
-        }
-    },
-    "Cooking spray": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Corded phone": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Corded phone/images",
-            "masks": "/openimages-mini/Corded phone/masks"
-        }
-    },
-    "Couch": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Couch/images",
-            "masks": "/openimages-mini/Couch/masks"
-        }
-    },
-    "Cowboy hat": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cowboy hat/images",
-            "masks": "/openimages-mini/Cowboy hat/masks"
-        }
-    },
-    "Cricket ball": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Crocodile": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Crocodile/images",
-            "masks": "/openimages-mini/Crocodile/masks"
-        }
-    },
-    "Croissant": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Croissant/images",
-            "masks": "/openimages-mini/Croissant/masks"
-        }
-    },
-    "Cucumber": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Cucumber/images",
-            "masks": "/openimages-mini/Cucumber/masks"
-        }
-    },
-    "Dagger": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dagger/images",
-            "masks": "/openimages-mini/Dagger/masks"
-        }
-    },
-    "Diaper": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Diaper/images",
-            "masks": "/openimages-mini/Diaper/masks"
-        }
-    },
-    "Dice": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dice/images",
-            "masks": "/openimages-mini/Dice/masks"
-        }
-    },
-    "Digital clock": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Digital clock/images",
-            "masks": "/openimages-mini/Digital clock/masks"
-        }
-    },
-    "Dog": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dog/images",
-            "masks": "/openimages-mini/Dog/masks"
-        }
-    },
-    "Dog bed": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dog bed/images",
-            "masks": "/openimages-mini/Dog bed/masks"
-        }
-    },
-    "Dolphin": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dolphin/images",
-            "masks": "/openimages-mini/Dolphin/masks"
-        }
-    },
-    "Door handle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Door handle/images",
-            "masks": "/openimages-mini/Door handle/masks"
-        }
-    },
-    "Doughnut": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Doughnut/images",
-            "masks": "/openimages-mini/Doughnut/masks"
-        }
-    },
-    "Dress": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Dress/images",
-            "masks": "/openimages-mini/Dress/masks"
-        }
-    },
-    "Drill (Tool)": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Drill (Tool)/images",
-            "masks": "/openimages-mini/Drill (Tool)/masks"
-        }
-    },
-    "Drink": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Drink/images",
-            "masks": "/openimages-mini/Drink/masks"
-        }
-    },
-    "Drinking straw": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Drinking straw/images",
-            "masks": "/openimages-mini/Drinking straw/masks"
-        }
-    },
-    "Duck": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Duck/images",
-            "masks": "/openimages-mini/Duck/masks"
-        }
-    },
-    "Dumbbell": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Eagle": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Eagle/images",
-            "masks": "/openimages-mini/Eagle/masks"
-        }
-    },
-    "Elephant": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Elephant/images",
-            "masks": "/openimages-mini/Elephant/masks"
-        }
-    },
-    "Envelope": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Envelope/images",
-            "masks": "/openimages-mini/Envelope/masks"
-        }
-    },
-    "Eraser": {
-        "count": 0,
-        "prefix": "an",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Facial tissue holder": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Falcon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Falcon/images",
-            "masks": "/openimages-mini/Falcon/masks"
-        }
-    },
-    "Fedora": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Fedora/images",
-            "masks": "/openimages-mini/Fedora/masks"
-        }
-    },
-    "Filing cabinet": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Filing cabinet/images",
-            "masks": "/openimages-mini/Filing cabinet/masks"
-        }
-    },
-    "Fire hydrant": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Fire hydrant/images",
-            "masks": "/openimages-mini/Fire hydrant/masks"
-        }
-    },
-    "Fish": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Fish/images",
-            "masks": "/openimages-mini/Fish/masks"
-        }
-    },
-    "Flag": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flag/images",
-            "masks": "/openimages-mini/Flag/masks"
-        }
-    },
-    "Flashlight": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flashlight/images",
-            "masks": "/openimages-mini/Flashlight/masks"
-        }
-    },
-    "Flower": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flower/images",
-            "masks": "/openimages-mini/Flower/masks"
-        }
-    },
-    "Flowerpot": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flowerpot/images",
-            "masks": "/openimages-mini/Flowerpot/masks"
-        }
-    },
-    "Flute": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flute/images",
-            "masks": "/openimages-mini/Flute/masks"
-        }
-    },
-    "Flying disc": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Flying disc/images",
-            "masks": "/openimages-mini/Flying disc/masks"
-        }
-    },
-    "Food processor": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Food processor/images",
-            "masks": "/openimages-mini/Food processor/masks"
-        }
-    },
-    "Football": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Football/images",
-            "masks": "/openimages-mini/Football/masks"
-        }
-    },
-    "Fox": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Fox/images",
-            "masks": "/openimages-mini/Fox/masks"
-        }
-    },
-    "Frog": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Frog/images",
-            "masks": "/openimages-mini/Frog/masks"
-        }
-    },
-    "Frying pan": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Frying pan/images",
-            "masks": "/openimages-mini/Frying pan/masks"
-        }
-    },
-    "Garden Asparagus": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Garden Asparagus/images",
-            "masks": "/openimages-mini/Garden Asparagus/masks"
-        }
-    },
-    "Giraffe": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Giraffe/images",
-            "masks": "/openimages-mini/Giraffe/masks"
-        }
-    },
-    "Girl": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Girl/images",
-            "masks": "/openimages-mini/Girl/masks"
-        }
-    },
-    "Glove": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Glove/images",
-            "masks": "/openimages-mini/Glove/masks"
-        }
-    },
-    "Goat": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Goat/images",
-            "masks": "/openimages-mini/Goat/masks"
-        }
-    },
-    "Goldfish": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Goldfish/images",
-            "masks": "/openimages-mini/Goldfish/masks"
-        }
-    },
-    "Golf ball": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Golf ball/images",
-            "masks": "/openimages-mini/Golf ball/masks"
-        }
-    },
-    "Goose": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Goose/images",
-            "masks": "/openimages-mini/Goose/masks"
-        }
-    },
-    "Grape": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Grape/images",
-            "masks": "/openimages-mini/Grape/masks"
-        }
-    },
-    "Grapefruit": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Grapefruit/images",
-            "masks": "/openimages-mini/Grapefruit/masks"
-        }
-    },
-    "Guacamole": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Guacamole/images",
-            "masks": "/openimages-mini/Guacamole/masks"
-        }
-    },
-    "Guitar": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Guitar/images",
-            "masks": "/openimages-mini/Guitar/masks"
-        }
-    },
-    "Hair dryer": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Hair spray": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Hamburger": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hamburger/images",
-            "masks": "/openimages-mini/Hamburger/masks"
-        }
-    },
-    "Hammer": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hammer/images",
-            "masks": "/openimages-mini/Hammer/masks"
-        }
-    },
-    "Hamster": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hamster/images",
-            "masks": "/openimages-mini/Hamster/masks"
-        }
-    },
-    "Hand dryer": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Handbag": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Handbag/images",
-            "masks": "/openimages-mini/Handbag/masks"
-        }
-    },
-    "Handgun": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Handgun/images",
-            "masks": "/openimages-mini/Handgun/masks"
-        }
-    },
-    "Harbor seal": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Harbor seal/images",
-            "masks": "/openimages-mini/Harbor seal/masks"
-        }
-    },
-    "Harmonica": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Harpsichord": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Harpsichord/images",
-            "masks": "/openimages-mini/Harpsichord/masks"
-        }
-    },
-    "Hat": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hat/images",
-            "masks": "/openimages-mini/Hat/masks"
-        }
-    },
-    "Heater": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Heater/images",
-            "masks": "/openimages-mini/Heater/masks"
-        }
-    },
-    "Hedgehog": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hedgehog/images",
-            "masks": "/openimages-mini/Hedgehog/masks"
-        }
-    },
-    "High heels": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/High heels/images",
-            "masks": "/openimages-mini/High heels/masks"
-        }
-    },
-    "Hippopotamus": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hippopotamus/images",
-            "masks": "/openimages-mini/Hippopotamus/masks"
-        }
-    },
-    "Horse": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Horse/images",
-            "masks": "/openimages-mini/Horse/masks"
-        }
-    },
-    "Hot dog": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Hot dog/images",
-            "masks": "/openimages-mini/Hot dog/masks"
-        }
-    },
-    "Human body": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Human body/images",
-            "masks": "/openimages-mini/Human body/masks"
-        }
-    },
-    "Human ear": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Human ear/images",
-            "masks": "/openimages-mini/Human ear/masks"
-        }
-    },
-    "Human mouth": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Human mouth/images",
-            "masks": "/openimages-mini/Human mouth/masks"
-        }
-    },
-    "Ipod": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Ipod/images",
-            "masks": "/openimages-mini/Ipod/masks"
-        }
-    },
-    "Jaguar (Animal)": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Jaguar (Animal)/images",
-            "masks": "/openimages-mini/Jaguar (Animal)/masks"
-        }
-    },
-    "Jeans": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Jeans/images",
-            "masks": "/openimages-mini/Jeans/masks"
-        }
-    },
-    "Jet ski": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Jet ski/images",
-            "masks": "/openimages-mini/Jet ski/masks"
-        }
-    },
-    "Jug": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Jug/images",
-            "masks": "/openimages-mini/Jug/masks"
-        }
-    },
-    "Juice": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Juice/images",
-            "masks": "/openimages-mini/Juice/masks"
-        }
-    },
-    "Kangaroo": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Kangaroo/images",
-            "masks": "/openimages-mini/Kangaroo/masks"
-        }
-    },
-    "Kettle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Kettle/images",
-            "masks": "/openimages-mini/Kettle/masks"
-        }
-    },
-    "Kitchen knife": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Kitchen knife/images",
-            "masks": "/openimages-mini/Kitchen knife/masks"
-        }
-    },
-    "Kite": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Kite/images",
-            "masks": "/openimages-mini/Kite/masks"
-        }
-    },
-    "Knife": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Knife/images",
-            "masks": "/openimages-mini/Knife/masks"
-        }
-    },
-    "Koala": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Koala/images",
-            "masks": "/openimages-mini/Koala/masks"
-        }
-    },
-    "Ladle": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Ladle/images",
-            "masks": "/openimages-mini/Ladle/masks"
-        }
-    },
-    "Laptop": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Laptop/images",
-            "masks": "/openimages-mini/Laptop/masks"
-        }
-    },
-    "Lemon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lemon/images",
-            "masks": "/openimages-mini/Lemon/masks"
-        }
-    },
-    "Leopard": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Leopard/images",
-            "masks": "/openimages-mini/Leopard/masks"
-        }
-    },
-    "Light switch": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Light switch/images",
-            "masks": "/openimages-mini/Light switch/masks"
-        }
-    },
-    "Lighthouse": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lighthouse/images",
-            "masks": "/openimages-mini/Lighthouse/masks"
-        }
-    },
-    "Limousine": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Limousine/images",
-            "masks": "/openimages-mini/Limousine/masks"
-        }
-    },
-    "Lion": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lion/images",
-            "masks": "/openimages-mini/Lion/masks"
-        }
-    },
-    "Lipstick": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lipstick/images",
-            "masks": "/openimages-mini/Lipstick/masks"
-        }
-    },
-    "Lizard": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lizard/images",
-            "masks": "/openimages-mini/Lizard/masks"
-        }
-    },
-    "Loveseat": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Loveseat/images",
-            "masks": "/openimages-mini/Loveseat/masks"
-        }
-    },
-    "Luggage and bags": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Luggage and bags/images",
-            "masks": "/openimages-mini/Luggage and bags/masks"
-        }
-    },
-    "Lynx": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Lynx/images",
-            "masks": "/openimages-mini/Lynx/masks"
-        }
-    },
-    "Magpie": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Magpie/images",
-            "masks": "/openimages-mini/Magpie/masks"
-        }
-    },
-    "Man": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Man/images",
-            "masks": "/openimages-mini/Man/masks"
-        }
-    },
-    "Mango": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mango/images",
-            "masks": "/openimages-mini/Mango/masks"
-        }
-    },
-    "Maracas": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Measuring cup": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Microwave oven": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Microwave oven/images",
-            "masks": "/openimages-mini/Microwave oven/masks"
-        }
-    },
-    "Milk": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Milk/images",
-            "masks": "/openimages-mini/Milk/masks"
-        }
-    },
-    "Miniskirt": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Miniskirt/images",
-            "masks": "/openimages-mini/Miniskirt/masks"
-        }
-    },
-    "Missile": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Missile/images",
-            "masks": "/openimages-mini/Missile/masks"
-        }
-    },
-    "Mixing bowl": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mixing bowl/images",
-            "masks": "/openimages-mini/Mixing bowl/masks"
-        }
-    },
-    "Mobile phone": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mobile phone/images",
-            "masks": "/openimages-mini/Mobile phone/masks"
-        }
-    },
-    "Monkey": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Monkey/images",
-            "masks": "/openimages-mini/Monkey/masks"
-        }
-    },
-    "Motorcycle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Motorcycle/images",
-            "masks": "/openimages-mini/Motorcycle/masks"
-        }
-    },
-    "Mouse": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mouse/images",
-            "masks": "/openimages-mini/Mouse/masks"
-        }
-    },
-    "Muffin": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Muffin/images",
-            "masks": "/openimages-mini/Muffin/masks"
-        }
-    },
-    "Mug": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mug/images",
-            "masks": "/openimages-mini/Mug/masks"
-        }
-    },
-    "Mule": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mule/images",
-            "masks": "/openimages-mini/Mule/masks"
-        }
-    },
-    "Mushroom": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Mushroom/images",
-            "masks": "/openimages-mini/Mushroom/masks"
-        }
-    },
-    "Nail (Construction)": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Orange": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Orange/images",
-            "masks": "/openimages-mini/Orange/masks"
-        }
-    },
-    "Ostrich": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Ostrich/images",
-            "masks": "/openimages-mini/Ostrich/masks"
-        }
-    },
-    "Otter": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Otter/images",
-            "masks": "/openimages-mini/Otter/masks"
-        }
-    },
-    "Oven": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Oven/images",
-            "masks": "/openimages-mini/Oven/masks"
-        }
-    },
-    "Owl": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Owl/images",
-            "masks": "/openimages-mini/Owl/masks"
-        }
-    },
-    "Oyster": {
-        "count": 1,
-        "prefix": "an",
-        "folder": {
-            "images": "/openimages-mini/Oyster/images",
-            "masks": "/openimages-mini/Oyster/masks"
-        }
-    },
-    "Pancake": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pancake/images",
-            "masks": "/openimages-mini/Pancake/masks"
-        }
-    },
-    "Panda": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Panda/images",
-            "masks": "/openimages-mini/Panda/masks"
-        }
-    },
-    "Paper cutter": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Paper towel": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Paper towel/images",
-            "masks": "/openimages-mini/Paper towel/masks"
-        }
-    },
-    "Parrot": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Parrot/images",
-            "masks": "/openimages-mini/Parrot/masks"
-        }
-    },
-    "Pastry": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pastry/images",
-            "masks": "/openimages-mini/Pastry/masks"
-        }
-    },
-    "Peach": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Peach/images",
-            "masks": "/openimages-mini/Peach/masks"
-        }
-    },
-    "Pear": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pear/images",
-            "masks": "/openimages-mini/Pear/masks"
-        }
-    },
-    "Pen": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pen/images",
-            "masks": "/openimages-mini/Pen/masks"
-        }
-    },
-    "Pencil case": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pencil case/images",
-            "masks": "/openimages-mini/Pencil case/masks"
-        }
-    },
-    "Penguin": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Penguin/images",
-            "masks": "/openimages-mini/Penguin/masks"
-        }
-    },
-    "Person": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Person/images",
-            "masks": "/openimages-mini/Person/masks"
-        }
-    },
-    "Piano": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Piano/images",
-            "masks": "/openimages-mini/Piano/masks"
-        }
-    },
-    "Picture frame": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Picture frame/images",
-            "masks": "/openimages-mini/Picture frame/masks"
-        }
-    },
-    "Pig": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pig/images",
-            "masks": "/openimages-mini/Pig/masks"
-        }
-    },
-    "Pillow": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pillow/images",
-            "masks": "/openimages-mini/Pillow/masks"
-        }
-    },
-    "Pitcher (Container)": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pitcher (Container)/images",
-            "masks": "/openimages-mini/Pitcher (Container)/masks"
-        }
-    },
-    "Pizza": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pizza/images",
-            "masks": "/openimages-mini/Pizza/masks"
-        }
-    },
-    "Pizza cutter": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Plastic bag": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Plastic bag/images",
-            "masks": "/openimages-mini/Plastic bag/masks"
-        }
-    },
-    "Platter": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Platter/images",
-            "masks": "/openimages-mini/Platter/masks"
-        }
-    },
-    "Polar bear": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Polar bear/images",
-            "masks": "/openimages-mini/Polar bear/masks"
-        }
-    },
-    "Pomegranate": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pomegranate/images",
-            "masks": "/openimages-mini/Pomegranate/masks"
-        }
-    },
-    "Popcorn": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Popcorn/images",
-            "masks": "/openimages-mini/Popcorn/masks"
-        }
-    },
-    "Potato": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Potato/images",
-            "masks": "/openimages-mini/Potato/masks"
-        }
-    },
-    "Power plugs and sockets": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Power plugs and sockets/images",
-            "masks": "/openimages-mini/Power plugs and sockets/masks"
-        }
-    },
-    "Pressure cooker": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pressure cooker/images",
-            "masks": "/openimages-mini/Pressure cooker/masks"
-        }
-    },
-    "Pretzel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pretzel/images",
-            "masks": "/openimages-mini/Pretzel/masks"
-        }
-    },
-    "Printer": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Printer/images",
-            "masks": "/openimages-mini/Printer/masks"
-        }
-    },
-    "Pumpkin": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Pumpkin/images",
-            "masks": "/openimages-mini/Pumpkin/masks"
-        }
-    },
-    "Punching bag": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Rabbit": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Rabbit/images",
-            "masks": "/openimages-mini/Rabbit/masks"
-        }
-    },
-    "Raccoon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Raccoon/images",
-            "masks": "/openimages-mini/Raccoon/masks"
-        }
-    },
-    "Racket": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Racket/images",
-            "masks": "/openimages-mini/Racket/masks"
-        }
-    },
-    "Radish": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Radish/images",
-            "masks": "/openimages-mini/Radish/masks"
-        }
-    },
-    "Raven": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Raven/images",
-            "masks": "/openimages-mini/Raven/masks"
-        }
-    },
-    "Red panda": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Red panda/images",
-            "masks": "/openimages-mini/Red panda/masks"
-        }
-    },
-    "Remote control": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Remote control/images",
-            "masks": "/openimages-mini/Remote control/masks"
-        }
-    },
-    "Reptile": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Reptile/images",
-            "masks": "/openimages-mini/Reptile/masks"
-        }
-    },
-    "Rhinoceros": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Rhinoceros/images",
-            "masks": "/openimages-mini/Rhinoceros/masks"
-        }
-    },
-    "Rocket": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Rocket/images",
-            "masks": "/openimages-mini/Rocket/masks"
-        }
-    },
-    "Roller skates": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Roller skates/images",
-            "masks": "/openimages-mini/Roller skates/masks"
-        }
-    },
-    "Rose": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Rose/images",
-            "masks": "/openimages-mini/Rose/masks"
-        }
-    },
-    "Rugby ball": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Rugby ball/images",
-            "masks": "/openimages-mini/Rugby ball/masks"
-        }
-    },
-    "Ruler": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Ruler/images",
-            "masks": "/openimages-mini/Ruler/masks"
-        }
-    },
-    "Sandwich": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sandwich/images",
-            "masks": "/openimages-mini/Sandwich/masks"
-        }
-    },
-    "Saucer": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Saucer/images",
-            "masks": "/openimages-mini/Saucer/masks"
-        }
-    },
-    "Saxophone": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Saxophone/images",
-            "masks": "/openimages-mini/Saxophone/masks"
-        }
-    },
-    "Scarf": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Scarf/images",
-            "masks": "/openimages-mini/Scarf/masks"
-        }
-    },
-    "Scissors": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Scissors/images",
-            "masks": "/openimages-mini/Scissors/masks"
-        }
-    },
-    "Screwdriver": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Screwdriver/images",
-            "masks": "/openimages-mini/Screwdriver/masks"
-        }
-    },
-    "Sculpture": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sculpture/images",
-            "masks": "/openimages-mini/Sculpture/masks"
-        }
-    },
-    "Sea lion": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sea lion/images",
-            "masks": "/openimages-mini/Sea lion/masks"
-        }
-    },
-    "Sea turtle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sea turtle/images",
-            "masks": "/openimages-mini/Sea turtle/masks"
-        }
-    },
-    "Seat belt": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Seat belt/images",
-            "masks": "/openimages-mini/Seat belt/masks"
-        }
-    },
-    "Segway": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Segway/images",
-            "masks": "/openimages-mini/Segway/masks"
-        }
-    },
-    "Shark": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Shark/images",
-            "masks": "/openimages-mini/Shark/masks"
-        }
-    },
-    "Sheep": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sheep/images",
-            "masks": "/openimages-mini/Sheep/masks"
-        }
-    },
-    "Shirt": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Shirt/images",
-            "masks": "/openimages-mini/Shirt/masks"
-        }
-    },
-    "Shorts": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Shorts/images",
-            "masks": "/openimages-mini/Shorts/masks"
-        }
-    },
-    "Shower": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Shower/images",
-            "masks": "/openimages-mini/Shower/masks"
-        }
-    },
-    "Skateboard": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Skateboard/images",
-            "masks": "/openimages-mini/Skateboard/masks"
-        }
-    },
-    "Skirt": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Skirt/images",
-            "masks": "/openimages-mini/Skirt/masks"
-        }
-    },
-    "Skull": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Skull/images",
-            "masks": "/openimages-mini/Skull/masks"
-        }
-    },
-    "Skunk": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Skunk/images",
-            "masks": "/openimages-mini/Skunk/masks"
-        }
-    },
-    "Skyscraper": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Skyscraper/images",
-            "masks": "/openimages-mini/Skyscraper/masks"
-        }
-    },
-    "Slow cooker": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Slow cooker/images",
-            "masks": "/openimages-mini/Slow cooker/masks"
-        }
-    },
-    "Snake": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Snake/images",
-            "masks": "/openimages-mini/Snake/masks"
-        }
-    },
-    "Snowmobile": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Snowmobile/images",
-            "masks": "/openimages-mini/Snowmobile/masks"
-        }
-    },
-    "Soap dispenser": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Soap dispenser/images",
-            "masks": "/openimages-mini/Soap dispenser/masks"
-        }
-    },
-    "Sock": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sock/images",
-            "masks": "/openimages-mini/Sock/masks"
-        }
-    },
-    "Sofa bed": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sofa bed/images",
-            "masks": "/openimages-mini/Sofa bed/masks"
-        }
-    },
-    "Sombrero": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sombrero/images",
-            "masks": "/openimages-mini/Sombrero/masks"
-        }
-    },
-    "Sparrow": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sparrow/images",
-            "masks": "/openimages-mini/Sparrow/masks"
-        }
-    },
-    "Spatula": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Spatula/images",
-            "masks": "/openimages-mini/Spatula/masks"
-        }
-    },
-    "Spoon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Spoon/images",
-            "masks": "/openimages-mini/Spoon/masks"
-        }
-    },
-    "Squash (Plant)": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Squash (Plant)/images",
-            "masks": "/openimages-mini/Squash (Plant)/masks"
-        }
-    },
-    "Squirrel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Squirrel/images",
-            "masks": "/openimages-mini/Squirrel/masks"
-        }
-    },
-    "Stapler": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": null,
-            "masks": null
-        }
-    },
-    "Starfish": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Starfish/images",
-            "masks": "/openimages-mini/Starfish/masks"
-        }
-    },
-    "Stop sign": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Stop sign/images",
-            "masks": "/openimages-mini/Stop sign/masks"
-        }
-    },
-    "Strawberry": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Strawberry/images",
-            "masks": "/openimages-mini/Strawberry/masks"
-        }
-    },
-    "Studio couch": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Studio couch/images",
-            "masks": "/openimages-mini/Studio couch/masks"
-        }
-    },
-    "Submarine sandwich": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Submarine sandwich/images",
-            "masks": "/openimages-mini/Submarine sandwich/masks"
-        }
-    },
-    "Suit": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Suit/images",
-            "masks": "/openimages-mini/Suit/masks"
-        }
-    },
-    "Suitcase": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Suitcase/images",
-            "masks": "/openimages-mini/Suitcase/masks"
-        }
-    },
-    "Sun hat": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sun hat/images",
-            "masks": "/openimages-mini/Sun hat/masks"
-        }
-    },
-    "Surfboard": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Surfboard/images",
-            "masks": "/openimages-mini/Surfboard/masks"
-        }
-    },
-    "Swan": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Swan/images",
-            "masks": "/openimages-mini/Swan/masks"
-        }
-    },
-    "Swim cap": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Swim cap/images",
-            "masks": "/openimages-mini/Swim cap/masks"
-        }
-    },
-    "Swimwear": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Swimwear/images",
-            "masks": "/openimages-mini/Swimwear/masks"
-        }
-    },
-    "Sword": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Sword/images",
-            "masks": "/openimages-mini/Sword/masks"
-        }
-    },
-    "Table tennis racket": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Table tennis racket/images",
-            "masks": "/openimages-mini/Table tennis racket/masks"
-        }
-    },
-    "Tablet computer": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tablet computer/images",
-            "masks": "/openimages-mini/Tablet computer/masks"
-        }
-    },
-    "Tank": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tank/images",
-            "masks": "/openimages-mini/Tank/masks"
-        }
-    },
-    "Tap": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tap/images",
-            "masks": "/openimages-mini/Tap/masks"
-        }
-    },
-    "Tart": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tart/images",
-            "masks": "/openimages-mini/Tart/masks"
-        }
-    },
-    "Taxi": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Taxi/images",
-            "masks": "/openimages-mini/Taxi/masks"
-        }
-    },
-    "Tea": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tea/images",
-            "masks": "/openimages-mini/Tea/masks"
-        }
-    },
-    "Teapot": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Teapot/images",
-            "masks": "/openimages-mini/Teapot/masks"
-        }
-    },
-    "Teddy bear": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Teddy bear/images",
-            "masks": "/openimages-mini/Teddy bear/masks"
-        }
-    },
-    "Tennis ball": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tennis ball/images",
-            "masks": "/openimages-mini/Tennis ball/masks"
-        }
-    },
-    "Tennis racket": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tennis racket/images",
-            "masks": "/openimages-mini/Tennis racket/masks"
-        }
-    },
-    "Tie": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tie/images",
-            "masks": "/openimages-mini/Tie/masks"
-        }
-    },
-    "Tiger": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tiger/images",
-            "masks": "/openimages-mini/Tiger/masks"
-        }
-    },
-    "Toaster": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Toaster/images",
-            "masks": "/openimages-mini/Toaster/masks"
-        }
-    },
-    "Toilet": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Toilet/images",
-            "masks": "/openimages-mini/Toilet/masks"
-        }
-    },
-    "Toilet paper": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Toilet paper/images",
-            "masks": "/openimages-mini/Toilet paper/masks"
-        }
-    },
-    "Tomato": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tomato/images",
-            "masks": "/openimages-mini/Tomato/masks"
-        }
-    },
-    "Torch": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Torch/images",
-            "masks": "/openimages-mini/Torch/masks"
-        }
-    },
-    "Tortoise": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Tortoise/images",
-            "masks": "/openimages-mini/Tortoise/masks"
-        }
-    },
-    "Towel": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Towel/images",
-            "masks": "/openimages-mini/Towel/masks"
-        }
-    },
-    "Toy": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Toy/images",
-            "masks": "/openimages-mini/Toy/masks"
-        }
-    },
-    "Traffic light": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Traffic light/images",
-            "masks": "/openimages-mini/Traffic light/masks"
-        }
-    },
-    "Traffic sign": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Traffic sign/images",
-            "masks": "/openimages-mini/Traffic sign/masks"
-        }
-    },
-    "Train": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Train/images",
-            "masks": "/openimages-mini/Train/masks"
-        }
-    },
-    "Trousers": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Trousers/images",
-            "masks": "/openimages-mini/Trousers/masks"
-        }
-    },
-    "Truck": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Truck/images",
-            "masks": "/openimages-mini/Truck/masks"
-        }
-    },
-    "Turkey": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Turkey/images",
-            "masks": "/openimages-mini/Turkey/masks"
-        }
-    },
-    "Turtle": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Turtle/images",
-            "masks": "/openimages-mini/Turtle/masks"
-        }
-    },
-    "Van": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Van/images",
-            "masks": "/openimages-mini/Van/masks"
-        }
-    },
-    "Vase": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Vase/images",
-            "masks": "/openimages-mini/Vase/masks"
-        }
-    },
-    "Vehicle registration plate": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Vehicle registration plate/images",
-            "masks": "/openimages-mini/Vehicle registration plate/masks"
-        }
-    },
-    "Volleyball (Ball)": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Volleyball (Ball)/images",
-            "masks": "/openimages-mini/Volleyball (Ball)/masks"
-        }
-    },
-    "Waffle": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Waffle/images",
-            "masks": "/openimages-mini/Waffle/masks"
-        }
-    },
-    "Washing machine": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Washing machine/images",
-            "masks": "/openimages-mini/Washing machine/masks"
-        }
-    },
-    "Waste container": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Waste container/images",
-            "masks": "/openimages-mini/Waste container/masks"
-        }
-    },
-    "Watch": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Watch/images",
-            "masks": "/openimages-mini/Watch/masks"
-        }
-    },
-    "Watermelon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Watermelon/images",
-            "masks": "/openimages-mini/Watermelon/masks"
-        }
-    },
-    "Whale": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Whale/images",
-            "masks": "/openimages-mini/Whale/masks"
-        }
-    },
-    "Wheel": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Wheel/images",
-            "masks": "/openimages-mini/Wheel/masks"
-        }
-    },
-    "Whiteboard": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Whiteboard/images",
-            "masks": "/openimages-mini/Whiteboard/masks"
-        }
-    },
-    "Wine": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Wine/images",
-            "masks": "/openimages-mini/Wine/masks"
-        }
-    },
-    "Winter melon": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Winter melon/images",
-            "masks": "/openimages-mini/Winter melon/masks"
-        }
-    },
-    "Wok": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Wok/images",
-            "masks": "/openimages-mini/Wok/masks"
-        }
-    },
-    "Woman": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Woman/images",
-            "masks": "/openimages-mini/Woman/masks"
-        }
-    },
-    "Woodpecker": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Woodpecker/images",
-            "masks": "/openimages-mini/Woodpecker/masks"
-        }
-    },
-    "Wrench": {
-        "count": 0,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Wrench/images",
-            "masks": "/openimages-mini/Wrench/masks"
-        }
-    },
-    "Zebra": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Zebra/images",
-            "masks": "/openimages-mini/Zebra/masks"
-        }
-    },
-    "Zucchini": {
-        "count": 1,
-        "prefix": "a",
-        "folder": {
-            "images": "/openimages-mini/Zucchini/images",
-            "masks": "/openimages-mini/Zucchini/masks"
-        }
-    }
-}
diff --git a/determined/run_atman.yml b/determined/run_atman.yml
deleted file mode 100755
index d68deb2..0000000
--- a/determined/run_atman.yml
+++ /dev/null
@@ -1,48 +0,0 @@
-name: atman-eval
-project: atman
-max_restarts: 2
-environment:
-  force_pull_image: true
-resources:
-  priority: 40
-  slots_per_trial: 1
-  devices:
-    - container_path: /dev/infiniband
-      host_path: /dev/infiniband
-      mode: mrw
-searcher:
-  name: grid
-  metric: not_needed
-  max_concurrent_trials: 66
-  max_length: 10000
-
-bind_mounts:
-entrypoint: python3 run_eval.py
-hyperparameters:
-  result_path:
-    type: categorical
-    vals:
-      - /det_cos_nolog_mult
-  conc_sup:
-    type: categorical
-    vals:
-      - .1
-      - .2
-      - .3
-      - .4
-      - .5
-      - .6
-      - .7
-      - .8
-      - .9
-      - .99
-      - None
-  sup_fact:
-    type: categorical
-    vals:
-      - .1,.2
-      - .3,.4
-      - .5,.6
-      - .7,.8
-      - .9
-      - 1.,0.
diff --git a/determined/run_atman_layers.yml b/determined/run_atman_layers.yml
deleted file mode 100755
index 6a00ef9..0000000
--- a/determined/run_atman_layers.yml
+++ /dev/null
@@ -1,43 +0,0 @@
-name: atman-eval-layers
-project: atman
-max_restarts: 2
-environment:
-  force_pull_image: true
-resources:
-  priority: 40
-  slots_per_trial: 1
-  devices:
-    - container_path: /dev/infiniband
-      host_path: /dev/infiniband
-      mode: mrw
-searcher:
-  name: grid
-  metric: not_needed
-  max_concurrent_trials: 8
-  max_length: 10000
-
-bind_mounts:
-entrypoint: python3 run_eval.py
-hyperparameters:
-  result_path:
-    type: categorical
-    vals:
-      - /det_cos_nolog_mult_layers
-  conc_sup:
-    type: categorical
-    vals:
-      - .7
-  sup_fact:
-    type: categorical
-    vals:
-      - .7
-  layers:
-    type: categorical
-    vals:
-      - [[0,1],[2,3],[4,5],[6,7]]
-      - [[8,9],[10,11],[12,13],[14,15]]
-      - [[16,17],[18,19],[20,21],[22,23]]
-      - [[24,25],[26,27],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27]]
-      - [[0,1,2,3],[0,1,2,3,4,5],[0,1,2,3,4,5,6,7],[0,1,2,3,4,5,6,7,8,9]]
-      - [[0,1,2,3,4,5,6,7,8,9,10,11],[0,1,2,3,4,5,6,7,8,9,10,11,12,13],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]]
-      - [[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23],[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25]]
diff --git a/determined/run_eval.py b/determined/run_eval.py
deleted file mode 100755
index 21e38be..0000000
--- a/determined/run_eval.py
+++ /dev/null
@@ -1,120 +0,0 @@
-from datetime import datetime
-from atman_magma.magma  import Magma
-from atman_magma.explainer import Explainer
-from atman_magma.logit_parsing import get_delta_cross_entropies
-from atman_magma.openimages_eval import run_eval
-
-
-from multimodal_explain_eval.dataloader import DataLoader
-from multimodal_explain_eval.utils import load_json_as_dict
-
-import  sys
-
-from determined_eval_wrapper import process_determined_hparams
-
-import os
-import determined as det
-
-
-startTime = datetime.now()
-
-
-info = det.get_cluster_info()
-if info is None:
-    hparams = {}
-else:
-    hparams = info.trial.hparams if info.task_type == "TRIAL" else {}
-params = process_determined_hparams(hparams)
-
-
-
-if __name__ == "__main__":
-    conc_sup_values = str(params.conc_sup).split(',')
-    suppression_factor_values = str(params.sup_fact).split(',')
-
-    metadata = load_json_as_dict('metadata.json')
-
-    keys_to_delete = []
-    for x in metadata.keys():
-        if metadata[x]['count'] == 0:
-            keys_to_delete.append(x)
-
-    for key in keys_to_delete:
-        del metadata[key]
-
-
-    dataloader = DataLoader(
-        metadata=metadata
-    )
-
-    num_images = 0
-    for key in metadata:
-        num_images += metadata[key]["count"]
-        # print(f"{key:<17}:", metadata[key]["count"])
-
-    print(f"total {num_images} pairs across {len(metadata)} classes")
-
-    num_total_explanations = (
-        len(conc_sup_values) *len(suppression_factor_values) * num_images
-    )
-    print(f"Will run a total of: {num_total_explanations} explanations")
-
-    print('output folder names:')
-
-    all_layers = [None]
-    if params.layers is not None and params.layers != "None":
-        all_layers = params.layers
-    # else:
-
-    for layers in all_layers:
-        for x in conc_sup_values:
-            for y in suppression_factor_values:
-                print(f"CONFIG {x} {y} // {len(conc_sup_values)} {len(suppression_factor_values)}")
-                if x == "None":
-                    conc_sup_value = None
-                else:
-                    conc_sup_value = float(x)
-                suppression_factor_value = float(y)
-                folder_name = f"{params.result_path}/con_sup_thres_{str(x) if x is not None else None}_suppression_factor_{str(y) if y is not None else None}_layer_{','.join([str(l) for l in layers]) if layers is not None else 'None'}"
-
-                print('loading model...')
-                model = Magma.from_checkpoint(
-                    checkpoint_path = "./mp_rank_00_model_states.pt",
-                    device = 'cuda'
-                )
-                model = model.eval()
-
-                folder_idx = 0
-
-                num_total_explanations = num_images
-                ex = Explainer(
-                    model = model,
-                    device = 'cuda',
-                    tokenizer = model.tokenizer,
-                    conceptual_suppression_threshold = conc_sup_value,
-                    suppression_factor = suppression_factor_value,
-                    layers=layers
-                )
-
-                run_eval(
-                    explainer = ex,
-                    metadata = metadata,
-                    dataloader = dataloader,
-                    logit_parsing_fn=get_delta_cross_entropies,
-                    output_folder = folder_name,
-                    max_batch_size = 144,
-                    text_prompt =  'This is a picture of ',
-                    use_lowercase_target=True,
-                    auto_decide_a_or_an=True,
-                    progress=True,
-                    square_outputs = False,
-                    num_total_explanations=num_total_explanations,
-                    prompt_explain_indices = [i for i in range(144)]
-                )
-                folder_idx += 1
-
-                print('eval complete :)')
-
-
-
-    print("ALL DONE")
diff --git a/startup-hook.sh b/startup-hook.sh
index 3199be6..12a5ecc 100755
--- a/startup-hook.sh
+++ b/startup-hook.sh
@@ -3,6 +3,5 @@ pip install deepspeed==0.6.0
 pip install typeguard==2.11.1
 pip install opencv-python-headless==4.2.0.34
 pip install ./magma
-pip install ./atman-open-images-eval
 pip install ./atman-magma
 pip install gdown==4.4.0 captum