Randomly select a window from a 3D array using Tensorflow Queue -
i have 3d numpy array data set axbxc
. how can apply tensorflow randomshufflequeue
implement random selection (dequeue
or dequeue_many
) window size axbxc
on first 2 dimension of data set?
a = 500 b = 600 c = 400 data = np.random.rand(a, b, c) q. tf.randomshufflequeue(capacity=???, shapes=???)
i want use queue instead of tf.random_crop
reduce overhead of cpu-gpu communication incurred placeholder
.
Comments
Post a Comment