├── README.md
└── convert_to_csv.py


/README.md:
--------------------------------------------------------------------------------
1 | # convert-dataset-images-to-csv-python
2 | This script can be used to convert the dataset images into CSV format.
3 | 


--------------------------------------------------------------------------------
/convert_to_csv.py:
--------------------------------------------------------------------------------
 1 | # dependencies
 2 | # OS modules
 3 | import os
 4 | # Pandas
 5 | import pandas as pd
 6 | # In-built time module
 7 | import time
 8 | # tqdm for progress bars
 9 | from tqdm import tqdm
10 | # Pillow Image Library
11 | from PIL import Image
12 | # Numpy module
13 | import numpy as np
14 | 
15 | # A list for column names of csv
16 | columnNames = list()
17 | # A column for label
18 | columnNames.append('label')
19 | # Other pixels column
20 | # replace 784 with your image size, here it is 28x28=784
21 | # iterate and build headers
22 | for i in range(784):
23 |     pixel = str(i)
24 |     columnNames.append(pixel)
25 | 
26 | # Create a Pandas dataframe for storing data
27 | train_data = pd.DataFrame(columns = columnNames)
28 | 
29 | # calculates the total number of images in the dataset initially 0
30 | num_images = 0
31 | 
32 | # iterate through every folder of the dataset
33 | for i in range(0, 58):
34 | 
35 |     # print messeage
36 |     print("Iterating: " + str(i) + " folder")
37 | 
38 |     # itreate through every image in the folder
39 |     # tqdm shows progress bar
40 |     for file in tqdm(os.listdir(str(i))):
41 |         # open image using PIL Image module
42 |         img = Image.open(os.path.join(str(i), file))
43 |         # resize to 28x28, replace with your size
44 |         img = img.resize((28, 28), Image.NEAREST)
45 |         # load image  
46 |         img.load()
47 |         # create a numpy array for image pixels
48 |         imgdata = np.asarray(img, dtype="int32")
49 |        
50 |         # temporary array to store pixel values
51 |         data = []
52 |         data.append(str(i))
53 |         for y in range(28):
54 |             for x in range(28):
55 |                 data.append(imgdata[x][y])
56 | 
57 |         # add the data row to training data dataframe
58 |         train_data.loc[num_images] = data
59 | 
60 |         # increment the number of images
61 |         num_images += 1
62 | 
63 | # write the dataframe to the CSV file
64 | train_data.to_csv("train_converted.csv", index=False)


--------------------------------------------------------------------------------