app.py

import sys

sys.path.append('./')

from typing import Tuple

import os
import cv2
import math
import torch
import random
import numpy as np
import argparse

import PIL
from PIL import Image

import diffusers
from diffusers.utils import load_image
from diffusers.models import ControlNetModel
from diffusers import LCMScheduler

from huggingface_hub import hf_hub_download

import insightface
from insightface.app import FaceAnalysis

from style_template import styles
from pipeline_stable_diffusion_xl_instantid_full import StableDiffusionXLInstantIDPipeline
from model_util import load_models_xl, get_torch_device, torch_gc

import gradio as gr

# global variable
MAX_SEED = np.iinfo(np.int32).max
device = get_torch_device()
dtype = torch.float16 if str(device).__contains__("cuda") else torch.float32
STYLE_NAMES = list(styles.keys())
DEFAULT_STYLE_NAME = "(No style)"

# Load face encoder
app = FaceAnalysis(name='antelopev2', root='./', providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
app.prepare(ctx_id=0, det_size=(640, 640))

# Path to InstantID models
face_adapter = f'./checkpoints/ip-adapter.bin'
controlnet_path = f'./checkpoints/ControlNetModel'

# Load pipeline
controlnet = ControlNetModel.from_pretrained(controlnet_path, torch_dtype=dtype)


def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8", enable_lcm_arg=False):
    if pretrained_model_name_or_path.endswith(
            ".ckpt"
    ) or pretrained_model_name_or_path.endswith(".safetensors"):
        scheduler_kwargs = hf_hub_download(
            repo_id="wangqixun/YamerMIX_v8",
            subfolder="scheduler",
            filename="scheduler_config.json",
        )

        (tokenizers, text_encoders, unet, _, vae) = load_models_xl(
            pretrained_model_name_or_path=pretrained_model_name_or_path,
            scheduler_name=None,
            weight_dtype=dtype,
        )

        scheduler = diffusers.EulerDiscreteScheduler.from_config(scheduler_kwargs)
        pipe = StableDiffusionXLInstantIDPipeline(
            vae=vae,
            text_encoder=text_encoders[0],
            text_encoder_2=text_encoders[1],
            tokenizer=tokenizers[0],
            tokenizer_2=tokenizers[1],
            unet=unet,
            scheduler=scheduler,
            controlnet=controlnet,
        ).to(device)

    else:
        pipe = StableDiffusionXLInstantIDPipeline.from_pretrained(
            pretrained_model_name_or_path,
            controlnet=controlnet,
            torch_dtype=dtype,
            safety_checker=None,
            feature_extractor=None,
        ).to(device)

        pipe.scheduler = diffusers.EulerDiscreteScheduler.from_config(pipe.scheduler.config)

    pipe.load_ip_adapter_instantid(face_adapter)
    # load and disable LCM
    pipe.load_lora_weights("latent-consistency/lcm-lora-sdxl")
    pipe.disable_lora()

    def toggle_lcm_ui(value):
        if value:
            return (
                gr.update(minimum=0, maximum=100, step=1, value=5),
                gr.update(minimum=0.1, maximum=20.0, step=0.1, value=1.5)
            )
        else:
            return (
                gr.update(minimum=5, maximum=100, step=1, value=30),
                gr.update(minimum=0.1, maximum=20.0, step=0.1, value=5)
            )

    def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
        if randomize_seed:
            seed = random.randint(0, MAX_SEED)
        return seed

    def remove_tips():
        return gr.update(visible=False)

    def get_example():
        case = [
            [
                './examples/yann-lecun_resize.jpg',
                "a man",
                "Snow",
                "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
            ],
            [
                './examples/musk_resize.jpeg',
                "a man",
                "Mars",
                "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
            ],
            [
                './examples/sam_resize.png',
                "a man",
                "Jungle",
                "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, gree",
            ],
            [
                './examples/schmidhuber_resize.png',
                "a man",
                "Neon",
                "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
            ],
            [
                './examples/kaifu_resize.png',
                "a man",
                "Vibrant Color",
                "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
            ],
        ]
        return case

    def run_for_examples(face_file, prompt, style, negative_prompt):
        return generate_image(face_file, None, prompt, negative_prompt, style, 30, 0.8, 0.8, 5, 42, False, True)

    def convert_from_cv2_to_image(img: np.ndarray) -> Image:
        return Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))

    def convert_from_image_to_cv2(img: Image) -> np.ndarray:
        return cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)

    def draw_kps(image_pil, kps, color_list=[(255, 0, 0), (0, 255, 0), (0, 0, 255), (255, 255, 0), (255, 0, 255)]):
        stickwidth = 4
        limbSeq = np.array([[0, 2], [1, 2], [3, 2], [4, 2]])
        kps = np.array(kps)

        w, h = image_pil.size
        out_img = np.zeros([h, w, 3])

        for i in range(len(limbSeq)):
            index = limbSeq[i]
            color = color_list[index[0]]

            x = kps[index][:, 0]
            y = kps[index][:, 1]
            length = ((x[0] - x[1]) ** 2 + (y[0] - y[1]) ** 2) ** 0.5
            angle = math.degrees(math.atan2(y[0] - y[1], x[0] - x[1]))
            polygon = cv2.ellipse2Poly((int(np.mean(x)), int(np.mean(y))), (int(length / 2), stickwidth), int(angle), 0,
                                       360, 1)
            out_img = cv2.fillConvexPoly(out_img.copy(), polygon, color)
        out_img = (out_img * 0.6).astype(np.uint8)

        for idx_kp, kp in enumerate(kps):
            color = color_list[idx_kp]
            x, y = kp
            out_img = cv2.circle(out_img.copy(), (int(x), int(y)), 10, color, -1)

        out_img_pil = Image.fromarray(out_img.astype(np.uint8))
        return out_img_pil

    def resize_img(input_image, max_side=1280, min_side=1024, size=None,
                   pad_to_max_side=False, mode=PIL.Image.BILINEAR, base_pixel_number=64):

        w, h = input_image.size
        if size is not None:
            w_resize_new, h_resize_new = size
        else:
            ratio = min_side / min(h, w)
            w, h = round(ratio * w), round(ratio * h)
            ratio = max_side / max(h, w)
            input_image = input_image.resize([round(ratio * w), round(ratio * h)], mode)
            w_resize_new = (round(ratio * w) // base_pixel_number) * base_pixel_number
            h_resize_new = (round(ratio * h) // base_pixel_number) * base_pixel_number
        input_image = input_image.resize([w_resize_new, h_resize_new], mode)

        if pad_to_max_side:
            res = np.ones([max_side, max_side, 3], dtype=np.uint8) * 255
            offset_x = (max_side - w_resize_new) // 2
            offset_y = (max_side - h_resize_new) // 2
            res[offset_y:offset_y + h_resize_new, offset_x:offset_x + w_resize_new] = np.array(input_image)
            input_image = Image.fromarray(res)
        return input_image

    def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str, str]:
        p, n = styles.get(style_name, styles[DEFAULT_STYLE_NAME])
        return p.replace("{prompt}", positive), n + ' ' + negative

    def generate_image(face_image_path, pose_image_path, prompt, negative_prompt, style_name, num_steps,
                       identitynet_strength_ratio, adapter_strength_ratio, guidance_scale, seed, enable_LCM,
                       enhance_face_region, progress=gr.Progress(track_tqdm=True)):
        if enable_LCM:
            pipe.enable_lora()
            pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
        else:
            pipe.disable_lora()
            pipe.scheduler = diffusers.EulerDiscreteScheduler.from_config(pipe.scheduler.config)

        if face_image_path is None:
            raise gr.Error(f"Cannot find any input face image! Please upload the face image")

        if prompt is None:
            prompt = "a person"

        # apply the style template
        prompt, negative_prompt = apply_style(style_name, prompt, negative_prompt)

        face_image = load_image(face_image_path)
        face_image = resize_img(face_image)
        face_image_cv2 = convert_from_image_to_cv2(face_image)
        height, width, _ = face_image_cv2.shape

        # Extract face features
        face_info = app.get(face_image_cv2)

        if len(face_info) == 0:
            raise gr.Error(f"Cannot find any face in the image! Please upload another person image")

        face_info = sorted(face_info, key=lambda x: (x['bbox'][2] - x['bbox'][0]) * (x['bbox'][3] - x['bbox'][1]))[
            -1]  # only use the maximum face
        face_emb = face_info['embedding']
        face_kps = draw_kps(convert_from_cv2_to_image(face_image_cv2), face_info['kps'])

        if pose_image_path is not None:
            pose_image = load_image(pose_image_path)
            pose_image = resize_img(pose_image)
            pose_image_cv2 = convert_from_image_to_cv2(pose_image)

            face_info = app.get(pose_image_cv2)

            if len(face_info) == 0:
                raise gr.Error(f"Cannot find any face in the reference image! Please upload another person image")

            face_info = face_info[-1]
            face_kps = draw_kps(pose_image, face_info['kps'])

            width, height = face_kps.size

        if enhance_face_region:
            control_mask = np.zeros([height, width, 3])
            x1, y1, x2, y2 = face_info["bbox"]
            x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
            control_mask[y1:y2, x1:x2] = 255
            control_mask = Image.fromarray(control_mask.astype(np.uint8))
        else:
            control_mask = None

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

        print("Start inference...")
        print(f"[Debug] Prompt: {prompt}, \n[Debug] Neg Prompt: {negative_prompt}")

        pipe.set_ip_adapter_scale(adapter_strength_ratio)
        images = pipe(
            prompt=prompt,
            negative_prompt=negative_prompt,
            image_embeds=face_emb,
            image=face_kps,
            control_mask=control_mask,
            controlnet_conditioning_scale=float(identitynet_strength_ratio),
            num_inference_steps=num_steps,
            guidance_scale=guidance_scale,
            height=height,
            width=width,
            generator=generator
        ).images

        return images[0], gr.update(visible=True)

    ### Description
    title = r"""
    <h1 align="center">InstantID: 1张照片，无需训练，秒级生成个人写真</h1>
    """

    description = r"""
    用户指南:<br>
    1. 上传人物图片。 对于多人图像，我们只会检测最大的脸部。 确保脸部不要太小，并且没有明显遮挡或模糊。
    2. （可选）上传另一个人的图像作为参考姿势。 如果没有上传，我们将使用第一张图像来提取姿势。 如果您在步骤1中使用了裁剪后的脸部，建议上传它以提取新的姿势。
    3. （可选）输入文本prompt提示词，就像所有文生图应用中所做的那样.
    4. 点击 <b>提交</b> 按钮开始定制.
    5. 分享美图给你的好友吧, enjoy😊!
    """

    tips = r"""
    ### 使用技巧
    1. 如果你对相似度不满意, 尝试增加"IdentityNet Strength"和"Adapter Strength"  
    2. 如果你觉得饱和度太高, 首先尝试降低"Adapter strength"。如果仍然太高, 降低"“"IdentityNet strength"
    3. 如果你觉得文本控制不符合预期, 降低"Adapter strength"
    """

    css = '''
    .gradio-container {width: 85% !important}
    '''
    with gr.Blocks(css=css) as demo:

        # description
        gr.Markdown(title)
        gr.Markdown(description)

        with gr.Row():
            with gr.Column():
                # upload face image
                face_file = gr.Image(label="上传人物图片", type="filepath")

                # optional: upload a reference pose image
                pose_file = gr.Image(label="上传参考pose图片(可选)", type="filepath")

                # prompt
                prompt = gr.Textbox(label="Prompt",
                                    info="给出简单的提示就足以实现良好的面部保真度",
                                    placeholder="A photo of a person",
                                    value="")

                submit = gr.Button("Submit", variant="primary")

                enable_LCM = gr.Checkbox(
                    label="使用LCM加快推理速度", value=enable_lcm_arg,
                    info="LCM可加快推理速度，但是生成图像的质量会变差",
                )
                style = gr.Dropdown(label="风格模版", choices=STYLE_NAMES, value=DEFAULT_STYLE_NAME)

                # strength
                identitynet_strength_ratio = gr.Slider(
                    label="IdentityNet strength (for fidelity)",
                    minimum=0,
                    maximum=1.5,
                    step=0.05,
                    value=0.80,
                )
                adapter_strength_ratio = gr.Slider(
                    label="Image adapter strength (for detail)",
                    minimum=0,
                    maximum=1.5,
                    step=0.05,
                    value=0.80,
                )

                with gr.Accordion(open=False, label="高级选项"):
                    negative_prompt = gr.Textbox(
                        label="Negative Prompt",
                        placeholder="low quality",
                        value="(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
                    )
                    num_steps = gr.Slider(
                        label="Number of sample steps",
                        minimum=20,
                        maximum=100,
                        step=1,
                        value=5 if enable_lcm_arg else 30,
                    )
                    guidance_scale = gr.Slider(
                        label="Guidance scale",
                        minimum=0.1,
                        maximum=10.0,
                        step=0.1,
                        value=0 if enable_lcm_arg else 5,
                    )
                    seed = gr.Slider(
                        label="Seed",
                        minimum=0,
                        maximum=MAX_SEED,
                        step=1,
                        value=42,
                    )
                    randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
                    enhance_face_region = gr.Checkbox(label="Enhance non-face region", value=True)

            with gr.Column():
                gallery = gr.Image(label="生成图像")
                usage_tips = gr.Markdown(label="使用技巧", value=tips, visible=False)

            submit.click(
                fn=remove_tips,
                outputs=usage_tips,
            ).then(
                fn=randomize_seed_fn,
                inputs=[seed, randomize_seed],
                outputs=seed,
                queue=False,
                api_name=False,
            ).then(
                fn=generate_image,
                inputs=[face_file, pose_file, prompt, negative_prompt, style, num_steps, identitynet_strength_ratio,
                        adapter_strength_ratio, guidance_scale, seed, enable_LCM, enhance_face_region],
                outputs=[gallery, usage_tips]
            )

            enable_LCM.input(fn=toggle_lcm_ui, inputs=[enable_LCM], outputs=[num_steps, guidance_scale], queue=False)

        gr.Examples(
            examples=get_example(),
            inputs=[face_file, prompt, style, negative_prompt],
            run_on_click=True,
            fn=run_for_examples,
            outputs=[gallery, usage_tips],
            cache_examples=True,
        )


    demo.queue().launch(server_name='0.0.0.0', share=True)


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--pretrained_model_name_or_path", type=str, default="wangqixun/YamerMIX_v8")
    parser.add_argument("--enable_LCM", type=bool, default=os.environ.get("ENABLE_LCM", False))

    args = parser.parse_args()

    main(args.pretrained_model_name_or_path, args.enable_LCM)