app.py

import gradio as gr
import os

import torch
from PIL import Image
from diffusers import (
    AutoencoderKL,
    DiffusionPipeline,
    # UNet2DConditionModel,
)

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

# if torch.cuda.is_available():
#     torch_dtype = torch.float16
# else:
torch_dtype = torch.float32

# pipe = DepthMasterPipeline.from_pretrained('eval', torch_dtype=torch_dtype)
# unet = UNet2DConditionModel.from_pretrained(os.path.join('eval', f'unet'))
# pipe = DepthMasterPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
# unet = UNet2DConditionModel.from_pretrained(model_repo_id, subfolder="unet", torch_dtype=torch_dtype)
# pipe.unet = unet
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)

# MAX_SEED = np.iinfo(np.int32).max
# MAX_IMAGE_SIZE = 1024


# @spaces.GPU #[uncomment to use ZeroGPU]
def infer(
    input_image,
    progress=gr.Progress(track_tqdm=True),
):
    # if randomize_seed:
    #     seed = random.randint(0, MAX_SEED)

    # generator = torch.Generator().manual_seed(seed)

    # image = pipe(
    #     prompt=prompt,
    #     negative_prompt=negative_prompt,
    #     guidance_scale=guidance_scale,
    #     num_inference_steps=num_inference_steps,
    #     width=width,
    #     height=height,
    #     generator=generator,
    # ).images[0]
    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 */
}
"""

with gr.Blocks(css=css) as demo:
    gr.Markdown("# DepthMaster")
    gr.Markdown("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.")
    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="input-image", interactive=True)
        with gr.Column():
            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("Compute Depth")

    # # 添加示例图片选择器
    # with gr.Row():
    #     gr.Markdown("### example images")
    # with gr.Row(elem_id="example-gallery"):
    #     example_gallery = gr.Gallery(
    #         label="",
    #         value=example_images,
    #         elem_id="example-gallery",
    #         show_label=False,
    #         interactive=True,
    #         columns=10
    #     )

    # 设置默认图片点击后的操作
    # example_gallery.select(
    #     fn=lambda img_path: img_path,  # 回调函数：返回选择的路径
    #     inputs=[],
    #     outputs=input_image  # 输出设置为 Input Image
    # )
    # example_gallery.click(
    #     fn=load_example,  # 选择图片的回调
    #     inputs=[example_gallery],  # 输入：用户点击的图片
    #     outputs=[input_image]  # 输出：更新 Input Image
    # )


    # 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
    )

# 启动 Gradio 应用
demo.launch()
        # with gr.Column(scale=45):
        #     img_in = gr.Image(type="pil")
        # with gr.Column(scale=45):
        #     img_out = 

        # with gr.Row():
        #     prompt = gr.Text(
        #         label="Prompt",
        #         show_label=False,
        #         max_lines=1,
        #         placeholder="Enter your prompt",
        #         container=False,
        #     )

        #     run_button = gr.Button("Run", scale=0, variant="primary")

        # result = gr.Image(label="Result", show_label=False)

        # with gr.Accordion("Advanced Settings", open=False):
        #     negative_prompt = gr.Text(
        #         label="Negative prompt",
        #         max_lines=1,
        #         placeholder="Enter a negative prompt",
        #         visible=False,
        #     )

        #     seed = gr.Slider(
        #         label="Seed",
        #         minimum=0,
        #         maximum=MAX_SEED,
        #         step=1,
        #         value=0,
        #     )

        #     randomize_seed = gr.Checkbox(label="Randomize seed", value=True)

            # with gr.Row():
            #     width = gr.Slider(
            #         label="Width",
            #         minimum=256,
            #         maximum=MAX_IMAGE_SIZE,
            #         step=32,
            #         value=1024,  # Replace with defaults that work for your model
            #     )

            #     height = gr.Slider(
            #         label="Height",
            #         minimum=256,
            #         maximum=MAX_IMAGE_SIZE,
            #         step=32,
            #         value=1024,  # Replace with defaults that work for your model
            #     )

            # with gr.Row():
            #     guidance_scale = gr.Slider(
            #         label="Guidance scale",
            #         minimum=0.0,
            #         maximum=10.0,
            #         step=0.1,
            #         value=0.0,  # Replace with defaults that work for your model
            #     )

            #     num_inference_steps = gr.Slider(
            #         label="Number of inference steps",
            #         minimum=1,
            #         maximum=50,
            #         step=1,
            #         value=2,  # Replace with defaults that work for your model
            #     )

    #     gr.Examples(examples=examples, inputs=[prompt])
    # gr.on(
    #     triggers=[run_button.click, prompt.submit],
    #     fn=infer,
    #     inputs=[
    #         prompt,
    #         negative_prompt,
    #         seed,
    #         randomize_seed,
    #         # width,
    #         # height,
    #         # guidance_scale,
    #         # num_inference_steps,
    #     ],
    #     outputs=[result, seed],
    # )

# if __name__ == "__main__":
#     demo.launch()