Tools for dealing with TFRecords
Project description
TFRecords
Tools for dealing with TFRecords
pip install tensorflow-records
import tensorflow as tf
import tf.records as tfr
image = tf.ones([1024, 768, 3], dtype=tf.uint8)
label = tf.constant('cat')
spec = tfr.spec(
image=tfr.Tensor([1024, 768, 3], dtype='int'),
label='string'
)
serialized = tfr.serialize(spec, image=x, label=y)
# b'\n+\n\x10\n\x05label\x12\x07\n\x05\n\x03cat\n\x17\n\x05image\x1...'
tfr.parse(spec).sample(tf.constant(serialized))
# { 'image': <tf.Tensor: shape=(1024, 768, 3) ...>, 'label': <tf.Tensor: shape=() ...> }
# or you can parse multiple at once
tfr.parse(spec).batch(tf.constant([serialized, serialized]))
# { 'image': <tf.Tensor: shape=(2, 1024, 768, 3) ...>, 'label': <tf.Tensor: shape=(2,) ...> }
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for tensorflow_records-0.1.2-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8f68276d35a8cd440ed4809460287d1433af212c55630b838eb14576dfa09820 |
|
MD5 | 890c4df41877f907629434e69516f62c |
|
BLAKE2b-256 | 1bfb64bf9ce16d1a47c1100a0f984bd337104a62436cff2e55d9c72a9c14efb4 |