test_proc_o.py

import torch
from PIL import Image
import torchvision.transforms.functional as F
from src.pix2pix_turbo import Pix2Pix_Turbo
import numpy as np

def process_sketch(sketch_path, output_path, prompt, val_r=0.4, seed=42):
    # Load the model
    model = Pix2Pix_Turbo("sketch_to_image_stochastic")
    
    # Set the seed for reproducibility
    torch.manual_seed(seed)

    # Load the sketch image
    image = Image.open(sketch_path).convert("RGB")
    
    # Convert the image to tensor and threshold it
    image_t = F.to_tensor(image) > 0.5
    
    # Prepare the input tensor
    with torch.no_grad():
        c_t = image_t.unsqueeze(0).cuda().float()
        B, C, H, W = c_t.shape
        
        # Create a random noise map
        noise = torch.randn((1, 4, H // 8, W // 8), device=c_t.device)
        
        # Call the Pix2Pix model
        output_image = model(c_t, prompt, deterministic=False, r=val_r, noise_map=noise)
    
    # Convert the output tensor to PIL image
    output_pil = F.to_pil_image(output_image[0].cpu() * 0.5 + 0.5)
    
    # Save the output image
    output_pil.save(output_path)
    print(f"Output image saved to {output_path}")

if __name__ == "__main__":
    sketch_path = "sketch.png"
    output_path = "output.png"
    prompt = ("a fantasy concept art of a magical castle in the sky, ")
    
    process_sketch(sketch_path, output_path, prompt)