Filter same tag images and save another folder at google drive using google colaboratory

Since the images are already labeled, better to use these labels to categorize the images. Inside the labels folder, there are JSON files containing image label data. You can fetch image names and relevant disaster types from JSON files.

"metadata": {
    "sensor": "GEOEYE01",
    "provider_asset_type": "GEOEYE01",
    "gsd": 2.0916247,
    "capture_date": "2018-09-20T16:04:41.000Z",
    "off_nadir_angle": 28.017313,
    "pan_resolution": 0.52282465,
    "sun_azimuth": 153.94543,
    "sun_elevation": 53.722378,
    "target_azimuth": 190.82309,
    "disaster": "hurricane-florence",
    "disaster_type": "flooding",
    "catalog_id": "1050010012411600",
    "original_width": 1024,
    "original_height": 1024,
    "width": 1024,
    "height": 1024,
    "id": "MjU0Njk0MQ.clApx1C8IcFymibsGi1JLu1eKhU",
    "img_name": "hurricane-florence_00000324_post_disaster.png"
  }

You can use the following code piece. It's written to copy an image to its relevant category folder (Ex: image with disaster_type 'fire' -> /categorized/fire/). Ultimately all the images will be categorized into separate folders.

from google.colab import drive 
import os 
import json 
import shutil

drive.mount('/content/drive')
# change paths according to yours
main_folder_path = "/content/drive/My Drive/Backup/train"
images_folder_path = main_folder_path+"/images"
labels_folder_path = main_folder_path+"/labels"
categorized_folder_path = "/content/drive/My Drive/Backup/categorized"

os.chdir(main_folder_path)
for json_filename in os.listdir(labels_folder_path):
  json_path = os.path.join(main_folder_path, "labels", json_filename)
  f = open(json_path, 'r')
  data = json.load(f)
  disaster_type = data["metadata"]["disaster_type"]
  img_name = data["metadata"]["img_name"]
  print("disaster:", disaster_type, "image:", img_name)
  f.close()
  img_filepath = os.path.join(main_folder_path, "images", img_name)
  category_folderpath = os.path.join(categorized_folder_path, disaster_type)
  if os.path.exists(img_filepath):
    if not os.path.exists(category_folderpath):
      os.mkdir(category_folderpath)
    shutil.copy(img_filepath, category_folderpath)

The dataset you posted has 6 types of natural disasters, hurricane, volcano, earthquake, flooding, tsunami, and wildfire. Each image file name contains one of these words, so you can easily filter images that are relevant to floods.

from google.colab import drive 
import os

i_path = "/content/drive/My Drive/images"
flood_dir = "/content/drive/My Drive/flood_images"

drive.mount('/content/drive')
os.chdir(i_path)  

for file_name in os.listdir():
  file_path = f"{i_path}/{file_name}"
  if "flooding" in file_name:
    s_p = os.path.join(i_path, file_path)
    d_p = os.path.join(flood_dir)
    !mv "$s_p" "$d_p"