- Trending Categories
Data Structure
Networking
RDBMS
Operating System
Java
MS Excel
iOS
HTML
CSS
Android
Python
C Programming
C++
C#
MongoDB
MySQL
Javascript
PHP
Physics
Chemistry
Biology
Mathematics
English
Economics
Psychology
Social Studies
Fashion Studies
Legal Studies
- Selected Reading
- UPSC IAS Exams Notes
- Developer's Best Practices
- Questions and Answers
- Effective Resume Writing
- HR Interview Questions
- Computer Glossary
- Who is Who
PyTorch – How to normalize an image with mean and standard deviation?
The Normalize() transform normalizes an image with mean and standard deviation. The torchvision.transforms module provides many important transforms that can be used to perform different types of manipulations on the image data.
Normalize() accepts only tensor images of any size. A tensor image is a torch tensor. A tensor image may have n number of channels. The Normalize() transform normalizes the tensor image for each channel.
As this transform supports only tensor image, the PIL images should be first converted to a torch tensor. And after applying Normalize() transform, we convert the normalized torch tensor to a PIL image.
Steps
We could use the following steps to normalize an image with mean and standard deviation −
Import the required libraries. In all the following examples, the required Python libraries are torch, Pillow, and torchvision. Make sure you have already installed them.
import torch import torchvision import torchvision.transforms as T from PIL import Image
Read the input image. The input image is a PIL image or a torch tensor. If the input image is PIL image, convert it to a torch tensor.
img = Image.open('sunset.jpg')
# convert image to torch tensor
imgTensor = T.ToTensor()(img)Define a transform to normalize the image with mean and standard deviation. Here, we use mean and std of the ImageNet dataset.
transform = T.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))
Apply the above-defined transform on the input image to normalize the image.
normalized_imgTensor = transform(imgTensor)
Convert the normalized tensor image to PIL image.
normalized_img = T.ToPILImage()(normalized_imgTensor)
Show the normalized image.
normalized _img.show()
Input Image
This image is used as the input file in all the following examples.
Example 1
The following Python program normalizes the input image to mean and standard deviation. We use the mean and standard deviation of ImageNet dataset.
# import required libraries
import torch
import torchvision.transforms as T
from PIL import Image
# Read the input image
img = Image.open('sunset.jpg')
# convert image to torch tensor
imgTensor = T.ToTensor()(img)
# define a transform to normalize the tensor
transform = T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
# normalize the converted tensor using above defined transform
normalized_imgTensor = transform(imgTensor)
# convert the normalized tensor to PIL image
normalized_img = T.ToPILImage()(normalized_imgTensor)
# display the normalized PIL image
normalized_img.show()Output
It will produce the following output −
.jpg)
Example 2
In this example, we define a Compose transform to perform three transformations.
Convert the PIL image to tensor image.
Normalize the tensor image.
Convert the normalized image tensor to PIL image.
# import required libraries
import torch
import torchvision.transforms as T
from PIL import Image
# read the input image
img = Image.open('sunset.jpg')
# define a transform to:
# convert the PIL image to tensor
# normalize the tensor
# convert the tensor to PIL image
transform = T.Compose([ T.ToTensor(), T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)), T.ToPILImage()])
# apply the above tensor on input image
img = transform(img)
img.show()Output
It will produce the following output −
.jpg)
5K+ Views
