Image Loading and Transformation for Style Transferring in PyTorch

by Online Tutorials Library July 22, 2022

Image loading and transformation for Style Transferring in PyTorch

After importing all the necessary libraries and adding VGG-19 to our device, we have to load images in the memory on which we want to apply for style transfer. We have a content image, and style image and the target image will be the combination of both these images. Not every image needs to have the same size or pixel. To make the images equal, we will also apply the image transformation process.

Image loading and transformation for Style Transferring in PyTorch

Image Loading

We have to load the content image and style image in memory so that we can perform the operation on that. The loading process plays a vital role in the style transferring process. We need images in memory and style transferring process will not be possible before the loading process.

Code:

  #defining a method with three parameters i.e. image location, maximum size and shape   def load_image(img_path,max_size=400,shape=None):  # Open the image, convert it into RGB and store in a variable       image=Image.open(img_path).convert(‘RGB’)      # comparing image size with the maximum size       if max(image.size)>max_size:        size=max_size      else:        size=max(image.size)      # checking for the image shape      if shape is not None:         size=shape      #Applying appropriate transformation to our image such as Resize, ToTensor and Normalization      in_transform=transforms.Compose([          transforms.Resize(size),          transforms.ToTensor(),          transforms.Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))])      #Calling in_transform with our image       image=in_transform(image).unsqueeze(0) #unsqueeze(0) is used to add extra layer of dimensionality to the image      #Returning image       return image  #Calling load_image() with our image and add it to our device  content=load_image(‘ab.jpg’).to(device)  style=load_image(‘abc.jpg’,shape=content.shape[-2:]).to(device)  

Image Conversion

Before importing our images, we need to convert our images from tensor as to numpy images to ensure the compatibility with the plot package. We have done this before with the familiar image_converts helper function which we have previously used in Image Transforms in Image Recognition.

  def im_convert(tensor):    image=tensor.cpu().clone().detach().numpy()    image=image.transpose(1,2,0)    image=image*np.array((0.5,0.5,0.5))+np.array((0.5,0.5,0.5))    image=image.clip(0,1)    return image  

If we run this helper method then it will generate the error. We have to remove the single dimensional entries from the shape of the image and from the shape of the array. So we will squeeze our image before the transpose method.

Plotting the images

Code:

  fig, (ax1,ax2)=plt.subplots(1,2,figsize=(20,10))  ax1.imshow(im_convert(content))  ax1.axis(‘off’)  ax2.imshow(im_convert(style))  ax2.axis(‘off’)  

When we run it on Google Colab Notebook, it will give us the expected output as:

Image loading and transformation for Style Transferring in PyTorch

Next TopicFeature Extraction for Style Transferring

Image Loading and Transformation for Style Transferring in PyTorch

Image loading and transformation for Style Transferring in PyTorch

Image Loading

Image Conversion

Plotting the images

Program to find the sum of Even Numbers

PyTorch Custom Module

You may also like