List image paths of multiple formats in a Kaggle Dataset

user_15142054 0
Viewed 67

How to form a list of images of multiple formats in a Kaggle dataset where the image paths are like? Working in Kaggle I wanted to convert the image paths into the list so that I can store and perform operations but couldn't find a proper image traversing Algo to give me the required list result.

Tree for the image is:

|-data
   |-images
        |-ID0
          |--- img4tgh4r3.jpg
          |--- img324633.png
          |
          .
          .
        |-ID1
        .
        .

I tried using ls -a but how do you convert this structure and save it into a data type to reuse it.

import os
  

path = "/"
dir_list = os.listdir(path)
  
print("Files and directories in '", path, "' :") 
  
# print the list
print(dir_list)

This only lists the directories but not all the image types.

kaggle data-science python
1 Answers

This can be done using either os or glob modules in Python. I would suggest using glob as it facilitates more functionality w.r.t filenames in various scenarios.

SAMPLE CODE:


import glob
from tqdm import tdqm

# The required file extensions
fetch_formats = ['png', 'jpg', 'jpeg']

# Declare an empty list for storing the file names
img_list = list()

# State the directory of interest
path = working_dir + "images/**/*."

# Fetch each type of file from the given directory
for ff in tqdm(fetch_formats, desc="Fetching the filenames"):
    img_list.extend(list(glob.glob(path+ff)))

print(f"\nTotal number of images: {len(img_list)}")

NOTE:

  • Usage of tqdm is for generating the progress bar and can be avoided
  • The *.png* would imply any filename ending with .png`
  • The dir\**\*.png would imply any sub-directory inside dir which contains files whose names end with .png

Check out the official documentation for more information

user_129083540
Related