Error for running, list index out of range

[andresvivancov]

Hello, sorry
Thanks for sharing!!
I was trying to use it but I had an error, could you help me, please?
I downloaded and I tested with the original mnist compressed and uncompressed and also with python 2 and python 3

python convert_mnist_to_png.py /home/andre/mnistX/in3/ /home/andresviv/mnistX/out/

Log:
andresviv@andresviv-ThinkPad-P50:/mnistX/ULTIMATE/01/mnist_png-master$ python convert_mnist_to_png.py /home/andresviv/mnistX/in3/ /home/andresviv/mnistX/out/
writing /home/andresviv/mnistX/out/training/0/0.png
writing /home/andresviv/mnistX/out/training/3/1.png
Traceback (most recent call last):
File "convert_mnist_to_png.py", line 68, in
path.join(output_path, dataset))
File "convert_mnist_to_png.py", line 47, in write_dataset
output_filename = path.join(output_dirs[label], str(i) + ".png")
IndexError: list index out of range
andresviv@andresviv-ThinkPad-P50:/mnistX/ULTIMATE/01/mnist_png-master$

Thank you

[Allicnam]

i have the same problem, would like to know how you figured it out!

Edit: i just solved it lol.

Solution (Windows 10):

1. Extract the four .gz mnist packages into your desktop (or whichever directory you're using)
2. In lines 15,16,18 and 19, change the second dash for a period in the path ( i.e. 'train-images.idx3-ubyte' instead of 'train-images-idx3-ubyte' )

Make sure you send the apropiate path in the arguments and don't change the file names

worked for me at least...

[Vi5HaL]

My solution :
I convert EMNIST balanced by just increase range of for loop at line :35( in write_dataset() > output_dirs
for i in range(47))

here Images are rotated

so to get proper images replace this code at line around 50( i m not sure see code)
w.write(h, data_i ) with
w.write(h, np.rot90(np.fliplr(data_i)))

and add import numpy as np at begining

[ankurjha7]

Instead of hardcoding value in the range one can also use len(set(labels))) . This will handle all the cases with different labels

[Vi5HaL]

Thank you :)

…

On Sun, 2 Sep 2018 8:40 PM ankurjha7, ***@***.***> wrote: Instead of hardcoding value in the range one can also use len(set(labels))) . This will handle all the cases with different labels — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AdDjcXa70HKYZgM3mgPhDIa7IqTEcCV0ks5uW_TdgaJpZM4N2eCh> .