def test_blend(dtype):
image1 = tf.constant(
[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]], dtype=dtype
)
image2 = tf.constant(
[
[255, 255, 255, 255],
[255, 255, 255, 255],
[255, 255, 255, 255],
[255, 255, 255, 255],
],
dtype=dtype,
)
blended = compose_ops.blend(image1, image2, 0.5).numpy()
np.testing.assert_equal(
blended,
[
[128, 128, 128, 128],
[128, 128, 128, 128],
[128, 128, 128, 128],
[128, 128, 128, 128],
],
)
image1 = np.random.randint(0, 255, (4, 4, 3), np.uint8)
image2 = np.random.randint(0, 255, (4, 4, 3), np.uint8)
blended = compose_ops.blend(
tf.convert_to_tensor(image1), tf.convert_to_tensor(image2), 0.35
).numpy()
expected = blend_np(image1, image2, 0.35)
np.testing.assert_equal(blended, expected)
def sharpness_image(image: TensorLike, factor: Number) -> tf.Tensor:
"""Implements Sharpness function from PIL using TF ops."""
orig_image = image
image_dtype = image.dtype
# SMOOTH PIL Kernel.
image = tf.cast(image, tf.float32)
kernel = (
tf.constant(
[[1, 1, 1], [1, 5, 1], [1, 1, 1]], dtype=tf.float32, shape=[3, 3, 1, 1]
)
/ 13.0
)
# Tile across channel dimension.
kernel = tf.tile(kernel, [1, 1, 3, 1])
strides = [1, 1, 1, 1]
degenerate = tf.nn.depthwise_conv2d(
image, kernel, strides, padding="VALID", dilations=[1, 1]
)
degenerate = tf.clip_by_value(degenerate, 0.0, 255.0)
degenerate = tf.cast(degenerate, image_dtype)
# For the borders of the resulting image, fill in the values of the
# original image.
mask = tf.ones_like(degenerate)
padded_mask = tf.pad(mask, [[0, 0], [1, 1], [1, 1], [0, 0]])
padded_degenerate = tf.pad(degenerate, [[0, 0], [1, 1], [1, 1], [0, 0]])
result = tf.where(tf.equal(padded_mask, 1), padded_degenerate, orig_image)
# Blend the final result.
blended = blend(result, orig_image, factor)
return tf.cast(blended, image_dtype)
def test_blend(dtype):
image1 = tf.constant(
[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]], dtype=dtype)
image2 = tf.constant(
[
[255, 255, 255, 255],
[255, 255, 255, 255],
[255, 255, 255, 255],
[255, 255, 255, 255],
],
dtype=dtype,
)
blended = compose_ops.blend(image1, image2, 0.5).numpy()
np.testing.assert_equal(
blended,
[
[128, 128, 128, 128],
[128, 128, 128, 128],
[128, 128, 128, 128],
[128, 128, 128, 128],
],
)
np.random.seed(0)
image1 = np.random.randint(0, 255, (3, 5, 5), np.uint8)
image2 = np.random.randint(0, 255, (3, 5, 5), np.uint8)
tf.random.set_seed(0)
factor = tf.random.uniform(shape=[],
maxval=1,
dtype=tf.dtypes.float32,
seed=0)
blended = compose_ops.blend(tf.convert_to_tensor(image1),
tf.convert_to_tensor(image2), factor).numpy()
expected = blend_np(image1, image2, factor.numpy())
np.testing.assert_equal(blended, expected)
assert blended.dtype == expected.dtype
