W3cubDocs

/TensorFlow 2.4

tf.image.stateless_random_brightness

Adjust the brightness of images by a random factor deterministically.

Equivalent to adjust_brightness() using a delta randomly picked in the interval [-max_delta, max_delta).

Guarantees the same results given the same seed independent of how many times the function is called, and independent of global seed settings (e.g. tf.random.set_seed).

Usage Example:

x = [[[1.0, 2.0, 3.0],
      [4.0, 5.0, 6.0]],
     [[7.0, 8.0, 9.0],
      [10.0, 11.0, 12.0]]]
seed = (1, 2)
tf.image.stateless_random_brightness(x, 0.2, seed)
<tf.Tensor: shape=(2, 2, 3), dtype=float32, numpy=
array([[[ 1.1376241,  2.1376243,  3.1376243],
        [ 4.1376243,  5.1376243,  6.1376243]],
       [[ 7.1376243,  8.137624 ,  9.137624 ],
        [10.137624 , 11.137624 , 12.137624 ]]], dtype=float32)>
Args
image An image or images to adjust.
max_delta float, must be non-negative.
seed A shape [2] Tensor, the seed to the random number generator. Must have dtype int32 or int64. (When using XLA, only int32 is allowed.)
Returns
The brightness-adjusted image(s).
Raises
ValueError if max_delta is negative.

© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r2.4/api_docs/python/tf/image/stateless_random_brightness