Data augmentation - Revision history

Davrot: /* torchvision.transforms.RandomApply(transforms, p=0.5) */

2025-10-21T14:23:30Z

torchvision.transforms.RandomApply(transforms, p=0.5)

← Older revision		Revision as of 14:23, 21 October 2025
Line 5:		Line 5:
	Initial Image:		Initial Image:
	[[File:Data augmentation test image.jpg\|center]]		[[File:Data augmentation test image.jpg\|center]]


	Photo by Udo Ernst		Photo by Udo Ernst
Line 214:		Line 213:
	plt.subplot(1, 2, 2)		plt.subplot(1, 2, 2)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 17.png]]<div class="figure">
	~~[[File:Image17.png\|image17]]~~
	</div>		</div>

Line 228:		Line 227:
	plt.subplot(1, 2, 2)		plt.subplot(1, 2, 2)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 18.png]]<div class="figure">
	~~[[File:Image18.png\|image18]]~~
	</div>		</div>

Line 237:		Line 236:
	new_image = random_erasing_transform(torch_image)		new_image = random_erasing_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 19.png]]<div class="figure">
	~~[[File:Image19.png\|image19]]~~
	</div>		</div>

Line 258:		Line 257:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 20.png]]<div class="figure">
	~~[[File:Image20.png\|image20]]~~
	</div>		</div>

Line 270:		Line 269:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 21.png]]<div class="figure">
	~~[[File:Image21.png\|image21]]~~
	</div>		</div>

Line 280:		Line 279:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 22.png]]<div class="figure">
	~~[[File:Image22.png\|image22]]~~
	</div>		</div>

Line 293:		Line 292:
	new_image = sequential_transform((torch_image * 255).type(dtype=torch.uint8))		new_image = sequential_transform((torch_image * 255).type(dtype=torch.uint8))
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight>Depending on the transformation used, I can be possible to just-in-time (jit) compile it.<syntaxhighlight lang="python">sequential_transform_jit = torch.jit.script(sequential_transform)</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>Depending on the transformation used, I can be possible to just-in-time (jit) compile it.<syntaxhighlight lang="python">sequential_transform_jit = torch.jit.script(sequential_transform)</syntaxhighlight>[[File:20 23.png]]<div class="figure">
	~~[[File:Image23.png\|image23]]~~
	</div>		</div>

Line 306:		Line 305:
	new_image = compose_transform((torch_image * 255).type(dtype=torch.uint8))		new_image = compose_transform((torch_image * 255).type(dtype=torch.uint8))
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 24.png]]<div class="figure">
	~~[[File:Image24.png\|image24]]~~
	</div>		</div>

Line 324:		Line 323:
	new_image = randomapply_transform((torch_image * 255).type(dtype=torch.uint8))		new_image = randomapply_transform((torch_image * 255).type(dtype=torch.uint8))
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 25.png]]<div class="figure">
	~~[[File:Image25.png\|image25]]~~
	</div>		</div>

Davrot at 10:01, 21 October 2025

2025-10-21T10:01:07Z

← Older revision		Revision as of 10:01, 21 October 2025
Line 4:		Line 4:

	Initial Image:		Initial Image:
			[[File:Data augmentation test image.jpg\|center]]

	~~[[File:Data_augmentation_test_image.jpg\|Initial Image]]~~ Photo by Udo Ernst
			Photo by Udo Ernst

	== Loading an example image (with opencv2) ==		== Loading an example image (with opencv2) ==
Line 35:		Line 37:
	=== Into PyTorch ===		=== Into PyTorch ===
	First we need to convert the np.ndarray into a suitable torch tensor<div class="figure">		First we need to convert the np.ndarray into a suitable torch tensor<div class="figure">
	[[File:~~Image2~~.png~~\|image2~~]]
	~~</div>~~<syntaxhighlight lang="python">import torch		</div>[[File:20 2.png]]<syntaxhighlight lang="python">import torch

	torch_image = torch.tensor(		torch_image = torch.tensor(
Line 51:		Line 53:
	new_image = pad_transform(torch_image)		new_image = pad_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 3.png]]<div class="figure">
	~~[[File:Image3.png\|image3]]~~
	</div>		</div>

Line 65:		Line 67:
	new_image = resize_transform(torch_image)		new_image = resize_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 4.png]]<div class="figure">
	~~[[File:Image4.png\|image4]]~~
	</div>		</div>

Line 73:		Line 75:
	new_image = center_crop_transform(torch_image)		new_image = center_crop_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 5.png]]<div class="figure">
	~~[[File:Image5.png\|image5]]~~
	</div>		</div>

Line 86:		Line 88:
	plt.imshow(np.moveaxis(new_image[i].detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image[i].detach().numpy(), 0, 2))

	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 6.png]]<div class="figure">
	~~[[File:Image6.png\|image6]]~~
	</div>		</div>

Line 97:		Line 99:
	new_image = gray_transform(torch_image)		new_image = gray_transform(torch_image)
	plt.imshow(new_image.squeeze().detach().numpy(), cmap="gray")		plt.imshow(new_image.squeeze().detach().numpy(), cmap="gray")
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 7.png]]<div class="figure">
	~~[[File:Image7.png\|image7]]~~
	</div>		</div>

Line 110:		Line 112:
	plt.subplot(2, 1, i)		plt.subplot(2, 1, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 14.png]]<div class="figure">
	~~[[File:Image14.png\|image14]]~~
	</div>		</div>

Line 126:		Line 128:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 8.png]]<div class="figure">
	~~[[File:Image8.png\|image8]]~~
	</div>		</div>

Line 136:		Line 138:
	new_image = gauss_transform(torch_image)		new_image = gauss_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 9.png]]<div class="figure">
	~~[[File:Image9.png\|image9]]~~
	</div>		</div>

Line 148:		Line 150:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 10.png]]<div class="figure">
	~~[[File:Image10.png\|image10]]~~
	</div>		</div>

Line 158:		Line 160:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 11.png]]<div class="figure">
	~~[[File:Image11.png\|image11]]~~
	</div>		</div>

Line 168:		Line 170:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 12.png]]<div class="figure">
	~~[[File:Image12.png\|image12]]~~
	</div>		</div>

Line 178:		Line 180:
	plt.subplot(3, 3, i)		plt.subplot(3, 3, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 13.png]]<div class="figure">
	~~[[File:Image13.png\|image13]]~~
	</div>		</div>

Line 191:		Line 193:
	plt.subplot(2, 2, i)		plt.subplot(2, 2, i)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 15.png]]<div class="figure">
	~~[[File:Image15.png\|image15]]~~
	</div>		</div>

Line 199:		Line 201:
	new_image = random_solarize_transform(torch_image)		new_image = random_solarize_transform(torch_image)
	plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))		plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
	plt.show()</syntaxhighlight><div class="figure">		plt.show()</syntaxhighlight>[[File:20 16.png]]<div class="figure">
	~~[[File:Image16.png\|image16]]~~
	</div>		</div>

Davrot: Created page with "What is available as data augmentation methods in torchvision? Questions to [mailto:davrot@uni-bremen.de David Rotermund] Initial Image: Initial Image Photo by Udo Ernst == Loading an example image (with opencv2) == Load it via [https://docs.opencv.org/4.5.3/d4/da8/group__imgcodecs.html#ga288b8b3da0892bd651fce07b3bbd3a56 cv2.imread( filename[, flags]) -> retval]import cv2 import matplotlib...."

2025-10-21T09:51:15Z

Created page with "What is available as data augmentation methods in torchvision? Questions to [mailto:davrot@uni-bremen.de David Rotermund] Initial Image: Initial Image Photo by Udo Ernst == Loading an example image (with opencv2) == Load it via [https://docs.opencv.org/4.5.3/d4/da8/group__imgcodecs.html#ga288b8b3da0892bd651fce07b3bbd3a56 cv2.imread( filename[, flags]) -> retval]<syntaxhighlight lang="python">import cv2 import matplotlib...."

New page

What is available as data augmentation methods in torchvision?

Questions to [mailto:davrot@uni-bremen.de David Rotermund]

Initial Image:

[[File:Data_augmentation_test_image.jpg|Initial Image]] Photo by Udo Ernst

== Loading an example image (with opencv2) ==
Load it via [https://docs.opencv.org/4.5.3/d4/da8/group__imgcodecs.html#ga288b8b3da0892bd651fce07b3bbd3a56 cv2.imread( filename[, flags]) -> retval]<syntaxhighlight lang="python">import cv2
import matplotlib.pyplot as plt

filename: str = "data_augmentation_test_image.jpg"

original_image = cv2.imread(filename)

plt.imshow(original_image)
plt.show()</syntaxhighlight>[[File:20 0.png]]<div class="figure">
</div>As you can see (not very well I might add) is that the color channels are wrong. But may be we want no color anyway ( options can be found [https://docs.opencv.org/4.5.3/d8/d6a/group__imgcodecs__flags.html#ga61d9b0126a3e57d9277ac48327799c80 here] ):<syntaxhighlight lang="python">original_image = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)

plt.imshow(original_image, cmap="gray")
plt.show()</syntaxhighlight>[[File:20 1.png]]<div class="figure">
</div><syntaxhighlight lang="python">import numpy as np

original_image = cv2.imread(filename, cv2.IMREAD_COLOR)

# "Convert" from BlueGreenRed (BGR) to RGB (RedGreenBlue)
# This is a flip in the third dimension.
original_image = np.flip(original_image, axis=2)
plt.imshow(original_image)
plt.show()</syntaxhighlight>

== Torchvision: A selection of transformations ==

=== Into PyTorch ===
First we need to convert the np.ndarray into a suitable torch tensor<div class="figure">
[[File:Image2.png|image2]]
</div><syntaxhighlight lang="python">import torch

torch_image = torch.tensor(
np.moveaxis(original_image.astype(dtype=np.float32) / 255.0, 2, 0)
)
print(torch_image.shape) # -> torch.Size([3, 1200, 1600])</syntaxhighlight>Note: For the following random opertions, we can control the random seed of torch via [https://pytorch.org/docs/stable/generated/torch.manual_seed.html torch.manual_seed(seed)].

Some example transformations from [https://pytorch.org/vision/stable/transforms.html torchvision]:

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Pad torchvision.transforms.Pad(padding, fill=0, padding_mode=‘constant’)] ===
<syntaxhighlight lang="python">import torchvision as tv

pad_transform = tv.transforms.Pad(padding=(50, 100), fill=0.5)
new_image = pad_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image3.png|image3]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomHorizontalFlip torchvision.transforms.RandomHorizontalFlip(p=0.5)] ===
Horizontally flip the given image randomly with a given probability.

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomVerticalFlip torchvision.transforms.RandomVerticalFlip(p=0.5)] ===
Vertically flip the given image randomly with a given probability.

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Pad torchvision.transforms.Resize(size, interpolation=<InterpolationMode.BILINEAR: ‘bilinear’>, max_size=None, antialias=None)] ===
The Resize transform resizes an image.<syntaxhighlight lang="python">resize_transform = tv.transforms.Resize(size=(50, 100))
new_image = resize_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image4.png|image4]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.CenterCrop torchvision.transforms.CenterCrop(size)] ===
The CenterCrop transform crops the given image at the center.<syntaxhighlight lang="python">center_crop_transform = tv.transforms.CenterCrop(size=(250, 200))
new_image = center_crop_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image5.png|image5]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.FiveCrop torchvision.transforms.FiveCrop(size)] ===
The FiveCrop transform crops the given image into four corners and the central crop.<syntaxhighlight lang="python">position = (1, 3, 7, 9, 5)
five_crop_transform = tv.transforms.FiveCrop(size=(250, 200))
new_image = five_crop_transform(torch_image)

for i, p in enumerate(position):
plt.subplot(3, 3, p)
plt.imshow(np.moveaxis(new_image[i].detach().numpy(), 0, 2))

plt.show()</syntaxhighlight><div class="figure">
[[File:Image6.png|image6]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Scale torchvision.transforms.TenCrop(size, vertical_flip=False)] ===
Crop the given image into four corners and the central crop plus the flipped version of these (horizontal flipping is used by default).

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Grayscale torchvision.transforms.Grayscale(num_output_channels=1)] ===
The Grayscale transform converts an image to grayscale.<syntaxhighlight lang="python">gray_transform = tv.transforms.Grayscale()
new_image = gray_transform(torch_image)
plt.imshow(new_image.squeeze().detach().numpy(), cmap="gray")
plt.show()</syntaxhighlight><div class="figure">
[[File:Image7.png|image7]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomGrayscale torchvision.transforms.RandomGrayscale(p=0.1)] ===
Randomly convert image to grayscale with a probability of p (default 0.1).

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomInvert torchvision.transforms.RandomInvert(p=0.5)] ===
Inverts the colors of the given image randomly with a given probability.<syntaxhighlight lang="python">random_invert_transform = tv.transforms.RandomInvert(p=0.5)
for i in range(1, 3):
new_image = random_invert_transform(torch_image)
plt.subplot(2, 1, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image14.png|image14]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Normalize torchvision.transforms.Normalize(mean, std, inplace=False)] ===
Normalize a tensor image with mean and standard deviation.

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomEqualize torchvision.transforms.RandomEqualize(p=0.5)] ===
Equalize the histogram of the given image randomly with a given probability.

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.ColorJitter torchvision.transforms.ColorJitter(brightness=0, contrast=0, saturation=0, hue=0)] ===
The ColorJitter transform randomly changes the brightness, saturation, and other properties of an image.<syntaxhighlight lang="python">color_jitter_transform = tv.transforms.ColorJitter(brightness=0.75, hue=0.5)
for i in range(1, 10):
new_image = color_jitter_transform(torch_image)
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image8.png|image8]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.GaussianBlur torchvision.transforms.GaussianBlur(kernel_size, sigma=(0.1, 2.0))] ===
The GaussianBlur transform performs gaussian blur transform on an image.

Note: Big kernel sizes are slow. (51,51) is rather big. Kernel size needs to be odd and positive.<syntaxhighlight lang="python">gauss_transform = tv.transforms.GaussianBlur(kernel_size=(101, 101), sigma=(0.1, 10))
new_image = gauss_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image9.png|image9]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomPerspective torchvision.transforms.RandomPerspective(distortion_scale=0.5, p=0.5, interpolation=<InterpolationMode.BILINEAR: ‘bilinear’>, fill=0)] ===
The RandomPerspective transform performs random perspective transform on an image.<syntaxhighlight lang="python">random_perspective_transform = tv.transforms.RandomPerspective(
distortion_scale=0.6, p=1.0
)
for i in range(1, 10):
new_image = random_perspective_transform(torch_image)
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image10.png|image10]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomRotation torchvision.transforms.RandomRotation(degrees, interpolation=<InterpolationMode.NEAREST: ‘nearest’>, expand=False, center=None, fill=0, resample=None)] ===
The RandomRotation transform rotates an image with random angle.<syntaxhighlight lang="python">random_rotation_transform = tv.transforms.RandomRotation(degrees=(0, 180))
for i in range(1, 10):
new_image = random_rotation_transform(torch_image)
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image11.png|image11]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomAffine torchvision.transforms.RandomAffine(degrees, translate=None, scale=None, shear=None, interpolation=<InterpolationMode.NEAREST: ‘nearest’>, fill=0, fillcolor=None, resample=None)] ===
The RandomAffine transform performs random affine transform on an image.<syntaxhighlight lang="python">random_affine_transform = tv.transforms.RandomAffine(degrees=(0, 180))
for i in range(1, 10):
new_image = random_affine_transform(torch_image)
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image12.png|image12]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomCrop torchvision.transforms.RandomCrop(size, padding=None, pad_if_needed=False, fill=0, padding_mode=‘constant’)] ===
The RandomCrop transform crops an image at a random location.<syntaxhighlight lang="python">random_crop_transform = tv.transforms.RandomCrop(size=(250, 200))
for i in range(1, 10):
new_image = random_crop_transform(torch_image)
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image13.png|image13]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomResizedCrop torchvision.transforms.RandomResizedCrop(size, scale=(0.08, 1.0), ratio=(0.75, 1.3333333333333333), interpolation=<InterpolationMode.BILINEAR: ‘bilinear’>)] ===
The RandomResizedCrop transform crops an image at a random location, and then resizes the crop to a given size.

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomPosterize torchvision.transforms.RandomPosterize(bits, p=0.5)] ===
Posterize the image randomly with a given probability by reducing the number of bits for each color channel.<syntaxhighlight lang="python">for i in range(1, 5):
random_posterize_transform = tv.transforms.RandomPosterize(bits=i, p=1.0)
new_image = random_posterize_transform((torch_image * 255).type(dtype=torch.uint8))
plt.subplot(2, 2, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image15.png|image15]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomSolarize torchvision.transforms.RandomSolarize(threshold, p=0.5)] ===
Solarize the image randomly with a given probability by inverting all pixel values above a threshold.<syntaxhighlight lang="python">random_solarize_transform = tv.transforms.RandomSolarize(threshold=0.5)
new_image = random_solarize_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image16.png|image16]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomAdjustSharpness torchvision.transforms.RandomAdjustSharpness(sharpness_factor, p=0.5)] ===
Adjust the sharpness of the image randomly with a given probability.<syntaxhighlight lang="python">random_sharpness_transform = tv.transforms.RandomAdjustSharpness(
sharpness_factor=50, p=1.0
)
new_image = random_sharpness_transform(torch_image)
plt.subplot(1, 2, 1)
plt.imshow(np.moveaxis(torch_image.detach().numpy(), 0, 2))
plt.subplot(1, 2, 2)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image17.png|image17]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomAutocontrast torchvision.transforms.RandomAutocontrast(p=0.5)] ===
Autocontrast the pixels of the given image randomly with a given probability.

I don’t see any effect.<syntaxhighlight lang="python">random_autocontrast_transform = tv.transforms.RandomAutocontrast(p=1.0)

new_image = random_autocontrast_transform(torch_image)
plt.subplot(1, 2, 1)
plt.imshow(np.moveaxis(torch_image.detach().numpy(), 0, 2))
plt.subplot(1, 2, 2)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image18.png|image18]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomErasing torchvision.transforms.RandomErasing(p=0.5, scale=(0.02, 0.33), ratio=(0.3, 3.3), value=0, inplace=False)] ===
Randomly selects a rectangle region in an torch Tensor image and erases its pixels.<syntaxhighlight lang="python">random_erasing_transform = tv.transforms.RandomErasing(p=1.0)

new_image = random_erasing_transform(torch_image)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image19.png|image19]]
</div>

== Predefined processing chains ==
[https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.AutoAugment torchvision.transforms.AutoAugment(policy: torchvision.transforms.autoaugment.AutoAugmentPolicy = <AutoAugmentPolicy.IMAGENET: ‘imagenet’>, interpolation: torchvision.transforms.functional.InterpolationMode = <InterpolationMode.NEAREST: ‘nearest’>, fill: Optional[List[float]] = None)]

AutoAugment data augmentation method based on [https://arxiv.org/pdf/1805.09501.pdf “AutoAugment: Learning Augmentation Strategies from Data”].

[https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.AutoAugmentPolicy torchvision.transforms.AutoAugmentPolicy(value)]

AutoAugment policies learned on different datasets. Available policies are IMAGENET, CIFAR10 and SVHN.

==== CIFAR10 ====
<syntaxhighlight lang="python">random_auto1_transform = tv.transforms.AutoAugment(
tv.transforms.AutoAugmentPolicy.CIFAR10
)
for i in range(1, 10):
new_image = random_auto1_transform((torch_image * 255).type(dtype=torch.uint8))
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image20.png|image20]]
</div>

==== IMAGENET ====
<syntaxhighlight lang="python">random_auto2_transform = tv.transforms.AutoAugment(
tv.transforms.AutoAugmentPolicy.IMAGENET
)
for i in range(1, 10):
new_image = random_auto2_transform((torch_image * 255).type(dtype=torch.uint8))
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image21.png|image21]]
</div>

==== SVHN ====
<syntaxhighlight lang="python">random_auto3_transform = tv.transforms.AutoAugment(tv.transforms.AutoAugmentPolicy.SVHN)
for i in range(1, 10):
new_image = random_auto3_transform((torch_image * 255).type(dtype=torch.uint8))
plt.subplot(3, 3, i)
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image22.png|image22]]
</div>

== Building custom processing chains ==

=== [https://pytorch.org/docs/stable/generated/torch.nn.Sequential.html#torch.nn.Sequential torch.nn.Sequential(*args)] ===
A sequential container. Modules will be added to it in the order they are passed in the constructor.<syntaxhighlight lang="python">sequential_transform = torch.nn.Sequential(
tv.transforms.RandomSolarize(threshold=0.5, p=1.0),
tv.transforms.RandomErasing(p=1.0),
)
new_image = sequential_transform((torch_image * 255).type(dtype=torch.uint8))
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight>Depending on the transformation used, I can be possible to just-in-time (jit) compile it.<syntaxhighlight lang="python">sequential_transform_jit = torch.jit.script(sequential_transform)</syntaxhighlight><div class="figure">
[[File:Image23.png|image23]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Compose torchvision.transforms.Compose(transforms)] ===
Composes several transforms together. '''This transform does not support torchscript.'''<syntaxhighlight lang="python">compose_transform = tv.transforms.Compose(
[
tv.transforms.RandomSolarize(threshold=0.5, p=1.0),
tv.transforms.RandomErasing(p=1.0),
]
)
new_image = compose_transform((torch_image * 255).type(dtype=torch.uint8))
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image24.png|image24]]
</div>

=== [https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.RandomApply torchvision.transforms.RandomApply(transforms, p=0.5)] ===
Apply randomly a list of transformations with a given probability.

'''Note: It randomly applies the whole list of transformation or none.'''<syntaxhighlight lang="python">randomapply_transform = tv.transforms.RandomApply(
[
tv.transforms.RandomSolarize(threshold=0.5, p=1.0),
tv.transforms.RandomErasing(p=1.0),
],
p=0.5,
)
for i in range(1, 3):
plt.subplot(2, 1, i)
new_image = randomapply_transform((torch_image * 255).type(dtype=torch.uint8))
plt.imshow(np.moveaxis(new_image.detach().numpy(), 0, 2))
plt.show()</syntaxhighlight><div class="figure">
[[File:Image25.png|image25]]
</div>

== Building your own filter ==
In the case you need a special filter then you just can write it very easily on your own. Here is an example.<syntaxhighlight lang="python">import torch

class OnOffFilter(torch.nn.Module):
def __init__(self, p: float = 0.5) -> None:
super(OnOffFilter, self).__init__()

self.p: float = p

def forward(self, tensor: torch.Tensor) -> torch.Tensor:

assert tensor.shape[1] == 1

tensor -= self.p
temp_0: torch.Tensor = torch.where(
tensor < 0.0, -tensor, tensor.new_zeros(tensor.shape, dtype=tensor.dtype)
)
temp_1: torch.Tensor = torch.where(
tensor >= 0.0, tensor, tensor.new_zeros(tensor.shape, dtype=tensor.dtype)
)

new_tensor: torch.Tensor = torch.cat((temp_0, temp_1), dim=1)

return new_tensor

def __repr__(self):
return self.__class__.__name__ + "(p={0})".format(self.p)

if __name__ == "__main__":
pass</syntaxhighlight>