RandomResizedCrop() Method in Python PyTorch

In this article, we are going to discuss RandomResizedCrop() method in Pytorch using Python.

RandomResizedCrop() method

RandomResizedCrop() method of torchvision.transforms module is used to crop a random area of the image and resized this image to the given size. This method accepts both PIL Image and Tensor Image. The tensor image is a PyTorch tensor with [C, H, W] shape, where C represents a number of channels and  H, W represents height and width respectively. This method returns a randomly cropped image. 

Syntax:  torchvision.transforms.RandomResizedCrop(size, scale, ratio)

Parameters:

• size: Desired crop size of the image.
• scale: This parameter is used to define the upper and lower bounds for the random area.
• ratio: This parameter is used to define upper and lower bounds for the random aspect ratio.

Return: This method will returns the randomly cropped image of given input size.

The below image is used for demonstration:

Example 1:

In this example, we are transforming the image with a height of 300 and a width of 600.

Python3

# import required libraries
import torch
import torchvision.transforms as transforms
from PIL import Image

# Read image
image = Image.open('pic.png')

# create an transform for crop the image
# 300px height and 600px wide
transform = transforms.RandomResizedCrop(size=(300, 600))

# use above created transform to crop
# the image
image_crop = transform(image)

# display result
image_crop.show()

Output:

Example 2:

In this example, we crop an image at a random location with the expected scale of 0.2 to 0.8.

Python3

# import required libraries
import torch
import torchvision.transforms as transforms
from PIL import Image

# Read image
image = Image.open('a.png')

# create an transform for crop the image
transform = transforms.RandomResizedCrop(size=(300, 600),
                                         scale=(0.2, 0.8))

# use above created transform to crop
# the image
image_crop = transform(image)

# display result
image_crop.show()

Output:

Example 3:

In this example, we crop an image at a random location with the expected ratio of 0.5 to 1.08.

Python3

# import required libraries
import torch
import torchvision.transforms as transforms
from PIL import Image

# Read image
image = Image.open('a.png')

# create an transform for crop the image
transform = transforms.RandomResizedCrop(
    size=(300, 600), scale=(0.2, 0.8), ratio=(0.5, 1.08))

# use above created transform to crop
# the image
image_crop = transform(image)

# display result
image_crop.show()

Output: