🚚 Move perlin noise to common folder (#424)

* Move perlin noise to common folder

* Fix linting issues

* Move file to data/utils + rename function

* Fix function names

* Fix mypy issue

* black

* fix type

Co-authored-by: Ashwin Vaidya <ashwinitinvaidya@gmail.com>

anomalib/data/utils/__init__.py

@@ -15,6 +15,7 @@
 # and limitations under the License.
 
 from .download import DownloadProgressBar, hash_check
+from .generators import random_2d_perlin
 from .image import get_image_filenames, read_image
 
-__all__ = ["get_image_filenames", "hash_check", "read_image", "DownloadProgressBar"]
+__all__ = ["get_image_filenames", "hash_check", "random_2d_perlin", "read_image", "DownloadProgressBar"]

anomalib/data/utils/generators/__init__.py

@@ -0,0 +1,8 @@
+"""Utilities to generate synthetic data."""
+
+# Copyright (C) 2022 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+from .perlin import random_2d_perlin
+
+__all__ = ["random_2d_perlin"]

anomalib/models/draem/utils/perlin.py → anomalib/data/utils/generators/perlin.py

@@ -12,9 +12,11 @@
 # pylint: disable=invalid-name
 
 import math
+from typing import Tuple, Union
 
 import numpy as np
 import torch
+from torch import Tensor
 
 
 def lerp_np(x, y, w):
@@ -63,7 +65,32 @@ def f(t):
     return np.sqrt(2) * ((1 - t[:, :, 1]) * n0 + t[:, :, 1] * n1)
 
 
-def rand_perlin_2d_np(shape, res, fade=lambda t: 6 * t**5 - 15 * t**4 + 10 * t**3):
+def random_2d_perlin(
+    shape: Tuple,
+    res: Tuple[Union[int, Tensor], Union[int, Tensor]],
+    fade=lambda t: 6 * t**5 - 15 * t**4 + 10 * t**3,
+) -> Union[np.ndarray, Tensor]:
+    """Returns a random 2d perlin noise array.
+
+    Args:
+        shape (Tuple): Shape of the 2d map.
+        res (Tuple[Union[int, Tensor]]): Tuple of scales for perlin noise for height and width dimension.
+        fade (_type_, optional): Function used for fading the resulting 2d map.
+            Defaults to equation 6*t**5-15*t**4+10*t**3.
+
+    Returns:
+        Union[np.ndarray, Tensor]: Random 2d-array/tensor generated using perlin noise.
+    """
+    if isinstance(res[0], int):
+        result = _rand_perlin_2d_np(shape, res, fade)
+    elif isinstance(res[0], Tensor):
+        result = _rand_perlin_2d(shape, res, fade)
+    else:
+        raise TypeError(f"got scales of type {type(res[0])}")
+    return result
+
+
+def _rand_perlin_2d_np(shape, res, fade=lambda t: 6 * t**5 - 15 * t**4 + 10 * t**3):
     """Generate a random image containing Perlin noise. Numpy version."""
     delta = (res[0] / shape[0], res[1] / shape[1])
     d = (shape[0] // res[0], shape[1] // res[1])
@@ -89,7 +116,7 @@ def dot(grad, shift):
     return math.sqrt(2) * lerp_np(lerp_np(n00, n10, t[..., 0]), lerp_np(n01, n11, t[..., 0]), t[..., 1])
 
 
-def rand_perlin_2d(shape, res, fade=lambda t: 6 * t**5 - 15 * t**4 + 10 * t**3):
+def _rand_perlin_2d(shape, res, fade=lambda t: 6 * t**5 - 15 * t**4 + 10 * t**3):
     """Generate a random image containing Perlin noise. PyTorch version."""
     delta = (res[0] / shape[0], res[1] / shape[1])
     d = (shape[0] // res[0], shape[1] // res[1])
@@ -128,7 +155,7 @@ def rand_perlin_2d_octaves(shape, res, octaves=1, persistence=0.5):
     frequency = 1
     amplitude = 1
     for _ in range(octaves):
-        noise += amplitude * rand_perlin_2d(shape, (frequency * res[0], frequency * res[1]))
+        noise += amplitude * _rand_perlin_2d(shape, (frequency * res[0], frequency * res[1]))
         frequency *= 2
         amplitude *= persistence
     return noise

anomalib/models/draem/utils/augmenter.py

@@ -21,7 +21,7 @@
 from torch import Tensor
 from torchvision.datasets.folder import IMG_EXTENSIONS
 
-from anomalib.models.draem.utils.perlin import rand_perlin_2d_np
+from anomalib.data.utils import random_2d_perlin
 
 
 class Augmenter:
@@ -84,7 +84,7 @@ def generate_perturbation(
         perlin_scalex = 2 ** random.randint(min_perlin_scale, perlin_scale)
         perlin_scaley = 2 ** random.randint(min_perlin_scale, perlin_scale)
 
-        perlin_noise = rand_perlin_2d_np((height, width), (perlin_scalex, perlin_scaley))
+        perlin_noise = random_2d_perlin((height, width), (perlin_scalex, perlin_scaley))
         perlin_noise = self.rot(image=perlin_noise)
 
         # Create mask from perlin noise