app.py

from __future__ import print_function
import torch
import torchvision
import torch.nn as nn
import gradio as gr
import os
import time
import torch.nn.functional as F
import torch.optim as optim
import matplotlib.pyplot as plt
import torchvision.transforms as transforms
import copy
import torchvision.models as models
import torchvision.transforms.functional as TF
from PIL import Image
import numpy as np
from model import ContentLoss, gram_matrix, StyleLoss, image_transform, get_input_optimizer,get_style_model_and_losses




#Defining the Model
cnn = models.vgg19(pretrained=True).features.eval()

#Normalization

cnn_normalization_mean = torch.tensor([0.485, 0.456, 0.406])
cnn_normalization_std = torch.tensor([0.229, 0.224, 0.225])


## style_transfer
import numpy as np
def run_style_transfer(cnn, normalization_mean, normalization_std,
                       content_img, style_img, input_img, num_steps=300,
                       style_weight=1000000, content_weight=1):
    """Run the style transfer."""
    print('Building the style transfer model..')
    model, style_losses, content_losses = get_style_model_and_losses(cnn,
        normalization_mean, normalization_std, style_img, content_img)

    # We want to optimize the input and not the model parameters so we
    # update all the requires_grad fields accordingly
    input_img.requires_grad_(True)
    model.requires_grad_(False)

    optimizer = get_input_optimizer(input_img)

    print('Optimizing..')
    run = [0]
    while run[0] <= num_steps:

        def closure():
            # correct the values of updated input image
            with torch.no_grad():
                input_img.clamp_(0, 1)

            optimizer.zero_grad()
            model(input_img)
            style_score = 0
            content_score = 0

            for sl in style_losses:
                style_score += sl.loss
            for cl in content_losses:
                content_score += cl.loss

            style_score *= style_weight
            content_score *= content_weight

            loss = style_score + content_score
            loss.backward()

            run[0] += 1
            if run[0] % 50 == 0:
                print("run {}:".format(run))
                print('Style Loss : {:4f} Content Loss: {:4f}'.format(
                    style_score.item(), content_score.item()))
                print()

            return style_score + content_score

        optimizer.step(closure)

    # a last correction...
    with torch.no_grad():
        input_img.clamp_(0, 1)

    # Convert output tensor to a NumPy array
    output_np = input_img.detach().cpu().numpy()[0].transpose(1, 2, 0)

    # Convert NumPy array to PIL Image object
    output_img = Image.fromarray((output_np * 255).astype(np.uint8))
    
    return output_img

# Define the dimensions of the random output image
image_width = 128
image_height = 128

#example_images
example_list = [["examples/" + example] for example in os.listdir("examples")]  

#Default output image
default_output_image = Image.fromarray(np.random.randint(0, 255, (image_height, image_width, 3), dtype=np.uint8))

#Defining the predict function
def style_transfer(cont_img:Image.Image=None,styl_img:Image.Image=None) -> Image.Image:

  if cont_img is None or styl_img is None:
        # Handle the default behavior when inputs are not provided
        # For example, you can load default images or return an error message
        # Return a default output image or error message
        return default_output_image
      
  #Start the timer
  start_time = time.time()
  
  #transform the input image
  style_img = image_transform(styl_img)
  content_img =image_transform(cont_img)
  print(styl_img)
  print(cont_img)
    
  #getting input image
  input_img = content_img.clone()

  #running the style transfer
  output = run_style_transfer(cnn, cnn_normalization_mean, cnn_normalization_std,
                            content_img, style_img, input_img)
  # output_img = output.detach().cpu().squeeze(0)
  # output_img = TF.to_pil_image(output_img)
  end_time=time.time()
  
  pred_time =round(end_time- start_time, 5)
  
  return output

##Gradio App
import gradio as gr
title= 'Style Transfer'
description='A model to transfer the style of one image to another'
article = 'Created at Pytorch Model Deployment'

#example_images
example_list = [["examples/" + example] for example in os.listdir("examples")]

#Create the gradio demo
demo = gr.Interface(
    fn=style_transfer,
    inputs=[
        gr.Image(label="content Image"),
        gr.Image(label="style_image")
    ],
    examples=example_list,
    outputs="image",
    allow_flagging=False,
    title=title,
    description=description,
    article=article
)

# Launch the Gradio interface
demo.launch(debug=True)