In this article, we will see how to apply a 2D average pooling in PyTorch.
AvgPool2d() method
AvgPool2d() method of torch.nn module is used to apply 2D average pooling over an input image composed of several input planes in PyTorch. The shape of the input 2D average pooling layer should be [N, C, H, W]. Where N represents the batch size, C represents the number of channels, and H, W represents the height and width of the input image respectively. The below syntax is used to apply 2D average pooling.
Syntax: torch.nn.AvgPool2d(kernel_size, stride)
Parameter:
- kernel_size: This is size of the window
- stride: This is stride of the window. The default value of stride is kernel_size.
Example 1:
Image used for demonstration:
In this example, we are applying 2D average pooling over an input image.
Python3
import torch from PIL import Image import torchvision.transforms as T image = Image.open('image1.png') # convert input image to torch tensor image = T.ToTensor()(image) # unsqueeze image to make 4D image = image.unsqueeze(0) # define avg pooling with square window # of (kernel_size=5, stride=3) pooling = torch.nn.AvgPool2d(5, 3) image = pooling(image) # squeeze image image = image.squeeze(0) # convert tensor to image image = T.ToPILImage()(image) image.show() |
Output:
Although the two images looks similar but if observed carefully then we can clearly say that the granularity of the image has been lost.
Example 2:
Image used for demonstration:
In this example, we are applying 2D average pooling over an input image.
Python3
import torch from PIL import Image import torchvision.transforms as T image = Image.open('image2.png') # convert input image to torch tensor image = T.ToTensor()(image) # unsqueeze image to make 4D image = image.unsqueeze(0) # define avg pooling with square window # of (kernel_size=5, stride=3) pooling = torch.nn.AvgPool2d(5, 3) image = pooling(image) # squeeze image image = image.squeeze(0) # convert tensor to image image = T.ToPILImage()(image) image.show() |
Output:
Here as well we can observe that the granularity of the image has been lost.
