app.py

import os
import cv2
import torch
import tempfile
import numpy as np
import matplotlib
import gradio as gr
from PIL import Image
import spaces
from gradio_imageslider import ImageSlider
from huggingface_hub import hf_hub_download
from bridge.dpt import Bridge  

# ====== Gradio CSS 样式 ======
css = """
#img-display-container {
    max-height: 100vh;
}
#img-display-input {
    max-height: 80vh;
}
#img-display-output {
    max-height: 80vh;
}
#download {
    height: 62px;
}
"""

# ====== device ====== 
DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'

# ====== model load ======
model = Bridge()
filepath = hf_hub_download(repo_id=f"Dingning/BRIDGE", filename=f"bridge.pth", repo_type="model")
state_dict = torch.load(filepath, map_location="cpu")


model.load_state_dict(state_dict)
model = model.to(DEVICE).eval()

# ====== description ======
title = "# Bridge Simplified Demo"
description = """
Official demo for Bridge using Gradio.  
[project page](https://dingning-liu.github.io/bridge.github.io/), 
[github](https://github.com/lnbxldn/BRIDGE).
"""

cmap = matplotlib.colormaps.get_cmap("Spectral_r")

# ====== inference ======
@spaces.GPU
def predict_depth(image: np.ndarray) -> np.ndarray:
    """Run depth inference on an RGB image (numpy)."""
    return model.infer_image(image[:, :, ::-1])  # BGR→RGB

def on_submit(image: np.ndarray):
    original_image = image.copy()
    depth = predict_depth(image)

    # 16-bit depth map
    raw_depth = Image.fromarray(depth.astype("uint16"))
    tmp_raw_depth = tempfile.NamedTemporaryFile(suffix=".png", delete=False)
    raw_depth.save(tmp_raw_depth.name)

    # normalization and colorize
    depth_norm = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
    depth_uint8 = depth_norm.astype(np.uint8)
    colored_depth = (cmap(depth_uint8)[:, :, :3] * 255).astype(np.uint8)

    # save depth map
    gray_depth = Image.fromarray(depth_uint8)
    tmp_gray_depth = tempfile.NamedTemporaryFile(suffix=".png", delete=False)
    gray_depth.save(tmp_gray_depth.name)

    return [(original_image, colored_depth), tmp_gray_depth.name, tmp_raw_depth.name]

# ====== Gradio UI======
with gr.Blocks(css=css) as demo:
    gr.Markdown(title)
    gr.Markdown(description)
    gr.Markdown("### Depth Prediction Demo")

    with gr.Row():
        input_image = gr.Image(
            label="Input Image", type="numpy", elem_id="img-display-input"
        )
        depth_image_slider = ImageSlider(
            label="Depth Map with Slider View", elem_id="img-display-output", position=0.5
        )
    submit = gr.Button(value="Compute Depth")
    gray_depth_file = gr.File(label="Grayscale depth map", elem_id="download")
    raw_file = gr.File(label="16-bit raw output", elem_id="download")

    submit.click(
        on_submit,
        inputs=[input_image],
        outputs=[depth_image_slider, gray_depth_file, raw_file]
    )

    # examples
    if os.path.exists("assets/examples"):
        example_files = sorted(os.listdir("assets/examples"))
        example_files = [os.path.join("assets/examples", f) for f in example_files]
        gr.Examples(
            examples=example_files,
            inputs=[input_image],
            outputs=[depth_image_slider, gray_depth_file, raw_file],
            fn=on_submit
        )

if __name__ == "__main__":
    demo.queue().launch(share=True)