CODEWITHSUNDEEP
pythontensorflow
Maggie
Maggie

Reputation: 357

Convert a list of images and labels to np array to train tensorflow

I'm using the tutorial about convolutional neural networks. In this function, I use:

# Train the model
train_input_fn = tf.estimator.inputs.numpy_input_fn(
    x={"x": X_train},
    y=y_train,
    batch_size=100,
    num_epochs=None,
    shuffle=True)
 mnist_classifier.train(
    input_fn=train_input_fn,
    steps=20000,
    hooks=[logging_hook])

where

type(X_train)

list

type(y_train)

list

y_train[0]

'0'

X_train[0].shape

(30, 29, 3)

type(X_train[0])

numpy.ndarray

len(X_train)

39209

len(y_train)

39209

I get the following error: AttributeError: 'list' object has no attribute 'shape'

Upvotes: 1

Views: 2419

Answers (1)

umutto
umutto

Reputation: 7700

X_train looks like a list of numpy arrays and tensorflow expects a numpy array, you can simply convert it to a numpy array by:

X_train = np.array(X_train)

or using numpy.asarray function, which does the exact same thing as above:

X_train = np.asarray(X_train)

Bear in mind, all your images should have the same dimensions for conversion to work.

Upvotes: 1

Related Questions