app.py

import gradio as gr
import numpy as np
import random
from PIL import Image
import torch
from diffusers import ControlNetModel, UniPCMultistepScheduler
from hico_pipeline import StableDiffusionControlNetMultiLayoutPipeline

device = "cuda" if torch.cuda.is_available() else "cpu"

# Initialize model
controlnet = ControlNetModel.from_pretrained("qihoo360/HiCo_T2I", torch_dtype=torch.float16)
pipe = StableDiffusionControlNetMultiLayoutPipeline.from_pretrained(
    "krnl/realisticVisionV51_v51VAE", controlnet=[controlnet], torch_dtype=torch.float16
)
pipe = pipe.to(device)
pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)

MAX_SEED = np.iinfo(np.int32).max

# Function for generating dummy bounding box and label data
def generate_dummy_data():
    # Generate random image size
    img_width, img_height = 512, 512
    r_image = np.zeros((img_height, img_width, 3), dtype=np.uint8)

    # Generate random bounding boxes and labels
    num_objects = random.randint(1, 5)
    r_obj_bbox = []
    r_obj_class = ["Object"]
    list_cond_image = []

    for _ in range(num_objects):
        x1, y1 = random.randint(0, img_width // 2), random.randint(0, img_height // 2)
        x2, y2 = random.randint(x1, img_width), random.randint(y1, img_height)
        r_obj_bbox.append([x1, y1, x2, y2])
        cond_image = np.zeros_like(r_image, dtype=np.uint8)
        cond_image[y1:y2, x1:x2] = 255
        list_cond_image.append(cond_image)

    r_obj_bbox.insert(0, [0, 0, img_width, img_height])  # Add background
    r_obj_class.insert(0, "Background")
    list_cond_image.insert(0, np.zeros_like(r_image, dtype=np.uint8))  # Add full background

    obj_cond_image = np.stack(list_cond_image, axis=0)
    list_cond_image_pil = [Image.fromarray(img).convert('RGB') for img in list_cond_image]

    return r_obj_class, r_obj_bbox, list_cond_image_pil, obj_cond_image

# Inference function
def infer(
    prompt, guidance_scale, num_inference_steps, randomize_seed, seed=None
):
    # Generate dummy data for demonstration
    r_obj_class, r_obj_bbox, list_cond_image_pil, _ = generate_dummy_data()
    if randomize_seed or seed is None:
        seed = random.randint(0, MAX_SEED)

    generator = torch.manual_seed(seed)

    # Run inference
    image = pipe(
        prompt=prompt,
        layo_prompt=r_obj_class,
        guess_mode=False,
        guidance_scale=guidance_scale,
        num_inference_steps=num_inference_steps,
        image=list_cond_image_pil,
        fuse_type="avg",
        width=512,
        height=512
    ).images[0]

    return image, seed

examples = [
    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
    "An astronaut riding a green horse",
    "A delicious ceviche cheesecake slice",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

# Gradio UI
with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(" # Text-to-Image Gradio Template")

        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):
            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():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=10.0,
                    step=0.1,
                    value=7.5,
                )

                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=50,
                )

        gr.Examples(examples=examples, inputs=[prompt])

    run_button.click(
        fn=infer,
        inputs=[
            prompt,
            guidance_scale,
            num_inference_steps,
            randomize_seed,
            seed,
        ],
        outputs=[result, seed],
    )

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