app.py

import gradio as gr
from DISTS_pytorch import DISTS
from torchvision.io import read_image
import torch
import torchvision.transforms.v2 as transforms
import spaces
from metrics.DeepDC import DeepDC
from metrics.DeepWSD import DeepWSD
from metrics.ADISTS import ADISTS
from dreamsim import dreamsim
# pyiqa requires older version of packages, causing dependency issues during install. Therefore, we install it here.
# Specifically, it requires transformers=4.37.2.
try:
    import pyiqa
except ImportError:
    print("pyiqa not found. Installing...")
    import subprocess
    import sys
    subprocess.check_call([sys.executable, "-m", "pip", "install", "pyiqa==0.1.14.1", "--no-deps"])
    import pyiqa

# Download models once at startup
_, _ = dreamsim(pretrained=True, device="cpu")

@spaces.GPU(duration=15)
class Evaluator:
    def __init__(self, device):
        self.device = device
        self.transform = transforms.ToDtype(dtype=torch.float32, scale=True)
        self.metrics = self._init_metrics()

    def _init_metrics(self):
        return {
            "↓ MSE": torch.nn.functional.mse_loss,
            "↓ L1": torch.nn.functional.l1_loss,
            "↓ DISTS": DISTS().to(self.device),
            "↓ ADISTS": ADISTS().to(self.device),            
            "↓ DeepDC": DeepDC().to(self.device),
            "↓ DeepWSD": DeepWSD().to(self.device),
            "↓ LPIPS": pyiqa.create_metric("lpips", device=self.device),
            "↓ DreamSim": dreamsim(pretrained=True, device=self.device)[0],
            "↑ PSNR": pyiqa.create_metric("psnr", device=self.device),
            "↑ SSIM": pyiqa.create_metric("ssim", device=self.device),
            "↑ MS-SSIM": pyiqa.create_metric("ms_ssim", device=self.device),
            "↑ CW-SSIM": pyiqa.create_metric("cw_ssim", device=self.device),
            "↑ FSIM": pyiqa.create_metric("fsim", device=self.device),
        }

    @torch.no_grad()
    def evaluate(self, img_fname1, img_fname2):
        img1 = self.transform(read_image(img_fname1)).unsqueeze(0).to(self.device)
        img2 = self.transform(read_image(img_fname2)).unsqueeze(0).to(self.device)

        # check images are the same size
        if img1.shape != img2.shape:
            return "Input images must have the same dimensions!"

        return "\n".join(
            f"{name:<10}: {float(metric(img1, img2).item()):3,.5f}"
            for name, metric in self.metrics.items()
        )


@spaces.GPU(duration=5)
def get_evaluator():
    """Returns a singleton Evaluator instance per worker/session."""
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    if not hasattr(get_evaluator, "evaluator"):
        get_evaluator.evaluator = Evaluator(device)
    return get_evaluator.evaluator


@spaces.GPU(duration=10)
def compute_similarity(img1_path, img2_path):
    """Main function for Gradio interface."""
    if not img1_path or not img2_path:
        return "Please upload both images!"
    return get_evaluator().evaluate(img1_path, img2_path)


def create_interface():
    examples = [
    ["examples/01_1.jpg", "examples/01_1.jpg"],  # Add an extra example for identical images
    ["examples/01_1.jpg", "examples/noise.jpg"],
    *[[f"examples/{i:02d}_1.jpg", f"examples/{i:02d}_2.jpg"] for i in range(1, 10)],
    ]

    # Custom CSS
    css = """
    .center-header {
        display: flex;
        align-items: center;
        justify-content: center;
        margin: 0 0 10px 0;
    }
    .monospace-text {
        font-family: 'Courier New', Courier, monospace;
    }
    .metrics-table {
        width: 100%;
        border-collapse: collapse;
    }
    .metrics-table td {
        padding: 10px;
        vertical-align: top;
    }    
    """

    # Add UI elements
    pyiqa_url = "https://github.com/chaofengc/IQA-PyTorch"
    with gr.Blocks(title="FR-IQA", css=css) as demo:
        gr.Markdown(f"""
        <div class='center-header'><h1>Full-Reference Image Quality Assessment</h1></div>
        
        Upload two images to compute various similarity metrics.<br>
        **Note**: Images must have identical dimensions. Code will run much faster locally: due to ZeroGPU setup, metrics are re-initialized on every run..
        """)

        with gr.Row():
            with gr.Column(scale=2):
                img_fname1 = gr.Image(type="filepath", label="Image#1", height=512, width=512)
            with gr.Column(scale=2):
                img_fname2 = gr.Image(type="filepath", label="Image#2", height=512, width=512)
            with gr.Column(scale=1):
                metrics_output = gr.Textbox(label="Metrics Output", lines=22, elem_classes="monospace-text", show_copy_button=True)

        with gr.Row():
            submit_btn = gr.Button("Compute Metrics", variant="primary")

        with gr.Row():
            with gr.Column(scale=2):
                gr.Examples(
                    examples=examples,
                    inputs=[img_fname1, img_fname2],
                    fn=compute_similarity,
                    outputs=metrics_output,
                    label="Example Pairs (all are 1024×768)",
                    cache_examples=True,
                    cache_mode="lazy",
                    examples_per_page=6
                )
            with gr.Column(scale=2):
                gr.Markdown(f"""
                <div class='center-header'><h3>Acknowledgements</h3></div>

                - Example images from [TryOffDiff](https://rizavelioglu.github.io/tryoffdiff) paper, which are sampled from VITON-HD dataset.
                - Metrics (*score range is only rough estimation, actual score range may vary*): 

                <table class="metrics-table">
                    <tr>
                        <th>Metric</th>
                        <th>Score Range</th>
                        <th>Lower is better?</th>
                        <th>Source</th>
                    </tr>
                    <tr>
                        <td>MSE</td>
                        <td>[0, ∞)</td>
                        <td>Yes</td>
                        <td><a href="https://docs.pytorch.org/docs/stable/generated/torch.nn.MSELoss.html">torch</a></td>
                    </tr>
                    <tr>
                        <td>L1</td>
                        <td>[0, ∞)</td>
                        <td>Yes</td>
                        <td><a href="https://docs.pytorch.org/docs/stable/generated/torch.nn.L1Loss.html">torch</a></td>
                    </tr>
                    <tr>
                        <td>DISTS</td>
                        <td>[0, 1]</td>
                        <td>Yes</td>
                        <td><a href="https://github.com/dingkeyan93/DISTS">official</a></td>
                    </tr>
                    <tr>
                        <td>ADISTS</td>
                        <td>~[0, 1]</td>
                        <td>Yes</td>
                        <td><a href="https://github.com/dingkeyan93/A-DISTS">official</a></td>
                    </tr>                
                    <tr>
                        <td>DeepDC</td>
                        <td>[0, 1]</td>
                        <td>Yes</td>
                        <td><a href="https://github.com/h4nwei/DeepDC">official</a></td>
                    </tr>
                    <tr>
                        <td>DeepWSD</td>
                        <td>[0, ∞)</td>
                        <td>Yes</td>
                        <td><a href="https://github.com/Buka-Xing/DeepWSD">official</a></td>
                    </tr>                
                    <tr>
                        <td>LPIPS</td>
                        <td>[0, 1]</td>
                        <td>Yes</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                    <tr>
                        <td>DreamSim</td>
                        <td>[0, 1]</td>
                        <td>Yes</td>
                        <td><a href="https://github.com/ssundaram21/dreamsim">official</a></td>
                    </tr>                    
                    <tr>
                        <td>PSNR</td>
                        <td>[0, ∞)</td>
                        <td>No</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                    <tr>
                        <td>SSIM</td>
                        <td>[0, 1]</td>
                        <td>No</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                    <tr>
                        <td>MS-SSIM</td>
                        <td>[0, 1]</td>
                        <td>No</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                    <tr>
                        <td>CW-SSIM</td>
                        <td>[0, 1]</td>
                        <td>No</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                    <tr>
                        <td>FSIM</td>
                        <td>[0, 1]</td>
                        <td>No</td>
                        <td><a href="{pyiqa_url}">pyiqa</a></td>
                    </tr>
                </table>
                """)

        submit_btn.click(
            fn=compute_similarity,
            inputs=[img_fname1, img_fname2],
            outputs=[metrics_output]
        )
    return demo


if __name__ == "__main__":
    demo = create_interface()
    demo.launch(share=False, ssr_mode=False)