Operation and visualization of Mnist dataset under TensorFlow __caffe&tensorflow&keras&theano

From tensorflow.examples.tutorials.mnist import Input_data

First you need to download the data set by networking:

Mnsit = Input_data.read_data_sets (train_dir= './mnist_data ', one_hot=true)
    # If there is no mnist_data under the current folder, the folder is created first, Then download the mnist dataset

Partition of training set and test set:

X_train, Y_train = Mnist.train.images, mnist.train.labels
        # returns X_train is a multidimensional array under NumPy, (55000, 784)
X_test, y_tes t = mnist.test.images, mnist.test.labels
        # (10000, 784)
x_valid, y_valid = Mnist.valid.images, Mnist.valid.labels
        # (5000, 784)

Of course, you can read the data in an iterative form with a certain batch_size:

Mnist.train.next_batch (100)

Mnist.train.next_batch () ⇒ Returns two values, one is the image data, one is the image data corresponding category information.

>> X_batch, Y_batch = mnist.train.next_batch (MB)
>> x_batch.shape
(784)
>> Y_ Batch.shape
(M) # One hot                 code

1. Visualization

# images:9* (28*28) 's Numpy.ndarray
# Y_ represents its true label information
def plot_mnist_3_3 (images, Y_, y=none):
    assert Images.shape[0] = = Len (y_)
    fig, axes = plt.subplots (3, 3)
    for I, Ax in Enumerate (Axes.flat):
        ax.imshow ( Images[i].reshape (image_shp), cmap= ' binary ')
        if Y is None:
            xlabel = ' True: {} '. Format (Y_[i])
        else:
            Xlabel = ' True: {0}, Pred: {1} '. Format (Y_[i], y[i])
        Ax.set_xlabel (Xlabel)
        ax.set_xticks ([])
        Ax.set_yticks ([])
    plt.show ()