CODEWITHSUNDEEP
tensorflow
user6819011
user6819011

Reputation:

Tensorflow: batching labels with tf.train.batch

i have a piece of code

la=[0,0,0,0,0,0,1,1,1,1]
onehot = tf.one_hot(la, depth=2)   #[[1,0],[1,0],[1,0],[1,0],[1,0],[1,0],[0,1],[0,1],[0,1],[0,1]]
image_batch,labels_batch=tf.train.batch([resized_image,onehot],batch_size=2,num_threads=1)

when i run

print(s.run([tf.shape(image_batch),labels_batch]))

it is batching all labes at a time,like

[array([ 2, 50, 50,  3]), array([[[ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 0.,  1.],
    [ 0.,  1.],
    [ 0.,  1.],
    [ 0.,  1.]],

   [[ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 1.,  0.],
    [ 0.,  1.],
    [ 0.,  1.],
    [ 0.,  1.],
    [ 0.,  1.]]], dtype=float32)]

it should output something like

[array([ 2, 50, 50,  3]), array([[[ 1.,  0.],

   [[ 1.,  0.]]], dtype=float32)]

doesn't it? as batch size is 2 and taking 2 images and it's corresponding labels at a time. i'm new to CNN and machine learning.thanks beforehand.

Upvotes: 0

Views: 387

Answers (1)

Nipun Wijerathne
Nipun Wijerathne

Reputation: 1829

According to the Tensorflow documentation of tf.train.batch (https://www.tensorflow.org/api_docs/python/tf/train/batch),

enter image description here

Since the enqueue_many=False by default and your input onehot have the shape of [10, 2], then the output (here labels_batch) shape become [batch_size, 10, 2].

if the enqueue_many=True, then only the output (here labels_batch) will become [batch_size,2].

Hope this helps.

Upvotes: 1

Related Questions