app.py

import gradio as gr
import os
import torch

import torch
from PIL import Image
from diffusers import (
    AutoencoderKL,
)

from transformers import CLIPTextModel, CLIPTokenizer
from depthmaster import DepthMasterPipeline
from depthmaster.modules.unet_2d_condition import UNet2DConditionModel

def load_example(example_image):
    # 返回选中的图片
    return example_image


device = "cuda" if torch.cuda.is_available() else "cpu"
model_repo_id = "zysong212/DepthMaster"  # Replace to the model you would like to use

torch_dtype = torch.float32

vae = AutoencoderKL.from_pretrained(model_repo_id, subfolder="vae", torch_dtype=torch_dtype, allow_pickle=False)
unet = UNet2DConditionModel.from_pretrained(model_repo_id, subfolder="unet", torch_dtype=torch_dtype, allow_pickle=False)
text_encoder = CLIPTextModel.from_pretrained(model_repo_id, subfolder="text_encoder", torch_dtype=torch_dtype)
tokenizer = CLIPTokenizer.from_pretrained(model_repo_id, subfolder="tokenizer", torch_dtype=torch_dtype)
pipe = DepthMasterPipeline(vae=vae, unet=unet, text_encoder=text_encoder, tokenizer=tokenizer)


try:
    pipe.enable_xformers_memory_efficient_attention()
except ImportError:
    pass  # run without xformers

pipe = pipe.to(device)


# @spaces.GPU #[uncomment to use ZeroGPU]
def infer(
    input_image,
    progress=gr.Progress(track_tqdm=True),
):

    pipe_out = pipe(
            input_image,
            processing_res=768,
            match_input_res=True,
            batch_size=1,
            color_map="Spectral",
            show_progress_bar=True,
            resample_method="bilinear",
        )

    # depth_pred: np.ndarray = pipe_out.depth_np
    depth_colored: Image.Image = pipe_out.depth_colored


    return depth_colored


# 默认图像路径
example_images = [
    "wild_example/000000000776.jpg",
    "wild_example/800x.jpg",
    "wild_example/000000055950.jpg",
    "wild_example/53441037037_c2cbd91ad2_k.jpg",
    "wild_example/53501906161_6109e3da29_b.jpg",
    "wild_example/m_1e31af1c.jpg",
    "wild_example/sg-11134201-7rd5x-lvlh48byidbqca.jpg"
]

# css = """
# #col-container {
#     margin: 0 auto;
#     max-width: 640px;
# }
# #example-gallery {
#     height: 80px; /* 设置缩略图高度 */
#     width: auto;  /* 保持宽高比 */
#     margin: 0 auto;  /* 图片间距 */
#     cursor: pointer; /* 鼠标指针变为手型 */
# }
# """

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

title = "# DepthMaster"
description = """**Official demo for DepthMaster**.
Please refer to our [paper](https://arxiv.org/abs/2501.02576), [project page](https://indu1ge.github.io/DepthMaster_page/), and [github](https://github.com/indu1ge/DepthMaster) for more details."""


with gr.Blocks(css=css) as demo:
    gr.Markdown(title)
    gr.Markdown(description)
    gr.Markdown(" ### Depth Estimation with DepthMaster.")
    # with gr.Column(elem_id="col-container"):
    #     gr.Markdown(" # Depth Estimation")
    with gr.Row():
        with gr.Column():
            input_image = gr.Image(label="Input Image", type="pil", elem_id="img-display-input")
        with gr.Column():
            # depth_img_slider = ImageSlider(label="Depth Map with Slider View", elem_id="img-display-output", position=0.5)
            depth_map = gr.Image(label="Depth Map with Slider View", type="pil", interactive=False, elem_id="depth-map")

    # 计算按钮
    compute_button = gr.Button(value="Compute Depth")

    # 设置计算按钮的回调
    compute_button.click(
        fn=infer,  # 回调函数
        inputs=[input_image],  # 输入
        outputs=[depth_map]  # 输出
    )

    example_files = os.listdir('wild_example')
    example_files.sort()
    example_files = [os.path.join('wild_example', filename) for filename in example_files]
    examples = gr.Examples(examples=example_files, inputs=[input_image], outputs=[depth_map], fn=infer)


# 启动 Gradio 应用
demo.queue().launch(share=True)