For reading datasets and converting to numpy files.

get_dataset.py

# Arda Mavi
import os
import numpy as np
from os import listdir
from scipy.misc import imread, imresize
from keras.utils import to_categorical
from sklearn.model_selection import train_test_split

# Settings:
img_size = 64
grayscale_images = True
num_class = 10
test_size = 0.2


def get_img(data_path):
    # Getting image array from path:
    img = imread(data_path, flatten=grayscale_images)
    img = imresize(img, (img_size, img_size, 1 if grayscale_images else 3))
    return img

def get_dataset(dataset_path='Dataset'):
    # Getting all data from data path:
    try:
        X = np.load('npy_dataset/X.npy')
        Y = np.load('npy_dataset/Y.npy')
    except:
        labels = listdir(dataset_path) # Geting labels
        X = []
        Y = []
        for i, label in enumerate(labels):
            datas_path = dataset_path+'/'+label
            for data in listdir(datas_path):
                img = get_img(datas_path+'/'+data)
                X.append(img)
                Y.append(i)
        # Create dateset:
        X = 1-np.array(X).astype('float32')/255.
        Y = np.array(Y).astype('float32')
        Y = to_categorical(Y, num_class)
        if not os.path.exists('npy_dataset/'):
            os.makedirs('npy_dataset/')
        np.save('npy_dataset/X.npy', X)
        np.save('npy_dataset/Y.npy', Y)
    X, X_test, Y, Y_test = train_test_split(X, Y, test_size=test_size, random_state=42)
    return X, X_test, Y, Y_test

if __name__ == '__main__':
    get_dataset()

Merhaba
Kodu calistirirken asagidaki uyariyi aliyorum:

/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/ipykernel_launcher.py:16: DeprecationWarning: `imread` is deprecated!
`imread` is deprecated in SciPy 1.0.0, and will be removed in 1.2.0.
Use ``imageio.imread`` instead.
app.launch_new_instance()

imread yerine imageio.imread tavsiye ediliyor.

Hi @ardamavi! Congratulations for the great job on making this dataset available!

I just would like to point out that you are assigning incorrect labels to the images.

Line 36 has to be changed from Y.append(i) to Y.append(int(label))

Best Regards,

@felipheggaliza

@hkarakose Yes, SciPy not support imread now. :( I back to OpenCv.

@felipheggalize Thank you but we can't use your method in Artificial Neural Networks. We need integer types for classes.

Please, fix the script. I was stuck the whole day trying to figure out why my model working wrong. The assigned labels differ from ground truth

Because list_dir() does not return file paths in a certain order, there might be a problem with the matching of images and labels as some users have noted. The origin of error is line 31 where enumerate() is applied in the unordered list of labels, resulting in the index i and label to be inconsistent. There are two solutions I can suggest:

1. The one referenced by @felipheggaliza which is to change line 36 from Y.append(i) to Y.append(int(label)). This also drops the necessity for using enumerate(), so you can change line 31 to: for label in labels: .

2. The other solution noted in @BenjaminFraser's fork where he sorts the output list of listdir() in line 28.

I think some of the functions are removed from the used libraries, so this is an updated version. Hope it works.

import os
import numpy as np
from os import listdir
from matplotlib.pyplot import imread
from skimage.transform import resize
from keras.utils import to_categorical
from sklearn.model_selection import train_test_split


# Settings:
img_size = 64
grayscale_images = True
num_class = 10
test_size = 0.2


def get_img(data_path):
    # Getting image array from path:
    img = imread(data_path)
    img = resize(img, (img_size, img_size, 1 if grayscale_images else 3))
    return img


def get_dataset(dataset_path='Dataset'):
    # Getting all data from data path:
    try:
        X = np.load('npy_dataset/X.npy')
        Y = np.load('npy_dataset/Y.npy')
    except:
        labels = listdir(dataset_path)  # Geting labels
        X = []
        Y = []
        for i, label in enumerate(labels):
            datas_path = dataset_path + '/' + label
            for data in listdir(datas_path):
                img = get_img(datas_path + '/' + data)
                X.append(img)
                Y.append(i)
        # Create dateset:
        X = 1 - np.array(X).astype('float32') / 255.
        Y = np.array(Y).astype('float32')
        Y = to_categorical(Y, num_class)
        if not os.path.exists('npy_dataset/'):
            os.makedirs('npy_dataset/')
        np.save('npy_dataset/X.npy', X)
        np.save('npy_dataset/Y.npy', Y)
    X, X_test, Y, Y_test = train_test_split(X, Y, test_size=test_size, random_state=42)
    return X, X_test, Y, Y_test


if __name__ == '__main__':
    X, X_test, Y, Y_test = get_dataset()
    print(X)
    print(X_test)
    print(Y)
    print(Y_test)

I think some of the functions are removed from the used libraries, so this is an updated version. Hope it works.

import os
import numpy as np
from os import listdir
from matplotlib.pyplot import imread
from skimage.transform import resize
from keras.utils import to_categorical
from sklearn.model_selection import train_test_split


# Settings:
img_size = 64
grayscale_images = True
num_class = 10
test_size = 0.2


def get_img(data_path):
    # Getting image array from path:
    img = imread(data_path)
    img = resize(img, (img_size, img_size, 1 if grayscale_images else 3))
    return img


def get_dataset(dataset_path='Dataset'):
    # Getting all data from data path:
    try:
        X = np.load('npy_dataset/X.npy')
        Y = np.load('npy_dataset/Y.npy')
    except:
        labels = listdir(dataset_path)  # Geting labels
        X = []
        Y = []
        for i, label in enumerate(labels):
            datas_path = dataset_path + '/' + label
            for data in listdir(datas_path):
                img = get_img(datas_path + '/' + data)
                X.append(img)
                Y.append(i)
        # Create dateset:
        X = 1 - np.array(X).astype('float32') / 255.
        Y = np.array(Y).astype('float32')
        Y = to_categorical(Y, num_class)
        if not os.path.exists('npy_dataset/'):
            os.makedirs('npy_dataset/')
        np.save('npy_dataset/X.npy', X)
        np.save('npy_dataset/Y.npy', Y)
    X, X_test, Y, Y_test = train_test_split(X, Y, test_size=test_size, random_state=42)
    return X, X_test, Y, Y_test


if __name__ == '__main__':
    X, X_test, Y, Y_test = get_dataset()
    print(X)
    print(X_test)
    print(Y)
    print(Y_test)

To make this code works fine, you should change the following lines:

1. Change the line "labels = listdir(dataset_path)" with "labels = ['0','1','2','3','4','5','6','7','8','9']"
2. Change the line "X = 1 - np.array(X).astype('float32') / 255." with "X = np.array(X).astype('float32')"

how to download this dataset " sign language digit dataset".
I tried this code on google colab, but it gave errors.
Please help

how to download this dataset " sign language digit dataset".
I tried this code on google colab, but it gave errors.
Please help

You can use Kaggle API to download and use the dataset.