mirror of https://github.com/InstantID/InstantID
Merge pull request #114 from InstantID/lcm_multicontrol
[Update Demo] Support LCM and Multi-ControlNetspull/141/head
InstantX
GitHub
commit
98332df4c1
|
|
@ -164,3 +164,5 @@ models/
|
|||
|
||||
# Cog
|
||||
.cog
|
||||
|
||||
gradio_cached_examples
|
||||
19
README.md
19
README.md
|
|
@ -7,8 +7,6 @@
|
|||
|
||||
<sup>*</sup>corresponding authors
|
||||
|
||||
{haofanwang.ai, wangqixun.ai}@gmail.com
|
||||
|
||||
<a href='https://instantid.github.io/'><img src='https://img.shields.io/badge/Project-Page-green'></a>
|
||||
<a href='https://arxiv.org/abs/2401.07519'><img src='https://img.shields.io/badge/Technique-Report-red'></a>
|
||||
<a href='https://huggingface.co/papers/2401.07519'><img src='https://img.shields.io/static/v1?label=Paper&message=Huggingface&color=orange'></a>
|
||||
|
|
@ -71,6 +69,13 @@ hf_hub_download(repo_id="InstantX/InstantID", filename="ControlNetModel/diffusio
|
|||
hf_hub_download(repo_id="InstantX/InstantID", filename="ip-adapter.bin", local_dir="./checkpoints")
|
||||
```
|
||||
|
||||
Or run the following command to download all models:
|
||||
|
||||
```python
|
||||
pip install -r gradio_demo/requirements.txt
|
||||
python gradio_demo/download_models.py
|
||||
```
|
||||
|
||||
If you cannot access to Huggingface, you can use [hf-mirror](https://hf-mirror.com/) to download models.
|
||||
```python
|
||||
export HF_ENDPOINT=https://hf-mirror.com
|
||||
|
|
@ -183,12 +188,17 @@ Run the following command:
|
|||
python gradio_demo/app.py
|
||||
```
|
||||
|
||||
or MultiControlNet version:
|
||||
```python
|
||||
gradio_demo/app-multicontrolnet.py
|
||||
```
|
||||
|
||||
## Usage Tips
|
||||
- For higher similarity, increase the weight of controlnet_conditioning_scale (IdentityNet) and ip_adapter_scale (Adapter).
|
||||
- For over-saturation, decrease the ip_adapter_scale. If not work, decrease controlnet_conditioning_scale.
|
||||
- For higher text control ability, decrease ip_adapter_scale.
|
||||
- For specific styles, choose corresponding base model makes differences.
|
||||
- We have not supported multi-person yet, will only use the largest face as reference pose.
|
||||
- We have not supported multi-person yet, only use the largest face as reference facial landmarks.
|
||||
- We provide a [style template](https://github.com/ahgsql/StyleSelectorXL/blob/main/sdxl_styles.json) for reference.
|
||||
|
||||
## Community Resources
|
||||
|
|
@ -218,9 +228,6 @@ python gradio_demo/app.py
|
|||
## Disclaimer
|
||||
The code of InstantID is released under [Apache License](https://github.com/InstantID/InstantID?tab=Apache-2.0-1-ov-file#readme) for both academic and commercial usage. **However, both manual-downloading and auto-downloading face models from insightface are for non-commercial research purposes only** accoreding to their [license](https://github.com/deepinsight/insightface?tab=readme-ov-file#license). Users are granted the freedom to create images using this tool, but they are obligated to comply with local laws and utilize it responsibly. The developers will not assume any responsibility for potential misuse by users.
|
||||
|
||||
## Declaration
|
||||
🚨🚨🚨 We solemnly clarify that [FAKE][FAKE][FAKE] http://instantid.org [FAKE][FAKE][FAKE] is not authorized and has no relationship with us. It is infringing and quite misleading, and has never contacted us for official cooperation. Please be aware of your personal privacy and subscription fraud. We reserve all legal rights.
|
||||
|
||||
## Star History
|
||||
|
||||
[](https://star-history.com/#InstantID/InstantID&Date)
|
||||
|
|
|
|||
Binary file not shown.
Binary file not shown.
|
After Width: | Height: | Size: 102 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 95 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 103 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 82 KiB |
|
|
@ -0,0 +1,674 @@
|
|||
import sys
|
||||
|
||||
sys.path.append("./")
|
||||
|
||||
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.pipelines.controlnet.multicontrolnet import MultiControlNetModel
|
||||
|
||||
from huggingface_hub import hf_hub_download
|
||||
|
||||
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
|
||||
from controlnet_util import openpose, get_depth_map, get_canny_image
|
||||
|
||||
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 = "Watercolor"
|
||||
|
||||
# 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 face ControlNetModel
|
||||
controlnet_identitynet = ControlNetModel.from_pretrained(
|
||||
controlnet_path, torch_dtype=dtype
|
||||
)
|
||||
|
||||
# controlnet-pose
|
||||
controlnet_pose_model = "thibaud/controlnet-openpose-sdxl-1.0"
|
||||
controlnet_canny_model = "diffusers/controlnet-canny-sdxl-1.0"
|
||||
controlnet_depth_model = "diffusers/controlnet-depth-sdxl-1.0-small"
|
||||
|
||||
controlnet_pose = ControlNetModel.from_pretrained(
|
||||
controlnet_pose_model, torch_dtype=dtype
|
||||
).to(device)
|
||||
controlnet_canny = ControlNetModel.from_pretrained(
|
||||
controlnet_canny_model, torch_dtype=dtype
|
||||
).to(device)
|
||||
controlnet_depth = ControlNetModel.from_pretrained(
|
||||
controlnet_depth_model, torch_dtype=dtype
|
||||
).to(device)
|
||||
|
||||
controlnet_map = {
|
||||
"pose": controlnet_pose,
|
||||
"canny": controlnet_canny,
|
||||
"depth": controlnet_depth,
|
||||
}
|
||||
controlnet_map_fn = {
|
||||
"pose": openpose,
|
||||
"canny": get_canny_image,
|
||||
"depth": get_depth_map,
|
||||
}
|
||||
|
||||
|
||||
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_identitynet],
|
||||
).to(device)
|
||||
|
||||
else:
|
||||
pipe = StableDiffusionXLInstantIDPipeline.from_pretrained(
|
||||
pretrained_model_name_or_path,
|
||||
controlnet=[controlnet_identitynet],
|
||||
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",
|
||||
None,
|
||||
"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",
|
||||
"./examples/poses/pose2.jpg",
|
||||
"a man flying in the sky in Mars",
|
||||
"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",
|
||||
"./examples/poses/pose4.jpg",
|
||||
"a man doing a silly pose wearing a suite",
|
||||
"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",
|
||||
"./examples/poses/pose3.jpg",
|
||||
"a man sit on a chair",
|
||||
"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",
|
||||
"./examples/poses/pose.jpg",
|
||||
"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, pose_file, prompt, style, negative_prompt):
|
||||
return generate_image(
|
||||
face_file,
|
||||
pose_file,
|
||||
prompt,
|
||||
negative_prompt,
|
||||
style,
|
||||
20, # num_steps
|
||||
0.8, # identitynet_strength_ratio
|
||||
0.8, # adapter_strength_ratio
|
||||
0.4, # pose_strength
|
||||
0.3, # canny_strength
|
||||
0.5, # depth_strength
|
||||
["pose", "canny"], # controlnet_selection
|
||||
5.0, # guidance_scale
|
||||
42, # seed
|
||||
"EulerDiscreteScheduler", # scheduler
|
||||
False, # enable_LCM
|
||||
True, # enable_Face_Region
|
||||
)
|
||||
|
||||
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,
|
||||
pose_strength,
|
||||
canny_strength,
|
||||
depth_strength,
|
||||
controlnet_selection,
|
||||
guidance_scale,
|
||||
seed,
|
||||
scheduler,
|
||||
enable_LCM,
|
||||
enhance_face_region,
|
||||
progress=gr.Progress(track_tqdm=True),
|
||||
):
|
||||
|
||||
if enable_LCM:
|
||||
pipe.scheduler = diffusers.LCMScheduler.from_config(pipe.scheduler.config)
|
||||
pipe.enable_lora()
|
||||
else:
|
||||
pipe.disable_lora()
|
||||
scheduler_class_name = scheduler.split("-")[0]
|
||||
|
||||
add_kwargs = {}
|
||||
if len(scheduler.split("-")) > 1:
|
||||
add_kwargs["use_karras_sigmas"] = True
|
||||
if len(scheduler.split("-")) > 2:
|
||||
add_kwargs["algorithm_type"] = "sde-dpmsolver++"
|
||||
scheduler = getattr(diffusers, scheduler_class_name)
|
||||
pipe.scheduler = scheduler.from_config(pipe.scheduler.config, **add_kwargs)
|
||||
|
||||
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, max_side=1024)
|
||||
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"Unable to detect a face in the image. Please upload a different photo with a clear face."
|
||||
)
|
||||
|
||||
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"])
|
||||
img_controlnet = face_image
|
||||
if pose_image_path is not None:
|
||||
pose_image = load_image(pose_image_path)
|
||||
pose_image = resize_img(pose_image, max_side=1024)
|
||||
img_controlnet = 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
|
||||
|
||||
if len(controlnet_selection) > 0:
|
||||
controlnet_scales = {
|
||||
"pose": pose_strength,
|
||||
"canny": canny_strength,
|
||||
"depth": depth_strength,
|
||||
}
|
||||
pipe.controlnet = MultiControlNetModel(
|
||||
[controlnet_identitynet]
|
||||
+ [controlnet_map[s] for s in controlnet_selection]
|
||||
)
|
||||
control_scales = [float(identitynet_strength_ratio)] + [
|
||||
controlnet_scales[s] for s in controlnet_selection
|
||||
]
|
||||
control_images = [face_kps] + [
|
||||
controlnet_map_fn[s](img_controlnet).resize((width, height))
|
||||
for s in controlnet_selection
|
||||
]
|
||||
else:
|
||||
pipe.controlnet = controlnet_identitynet
|
||||
control_scales = float(identitynet_strength_ratio)
|
||||
control_images = face_kps
|
||||
|
||||
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=control_images,
|
||||
control_mask=control_mask,
|
||||
controlnet_conditioning_scale=control_scales,
|
||||
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: Zero-shot Identity-Preserving Generation in Seconds</h1>
|
||||
"""
|
||||
|
||||
description = r"""
|
||||
<b>Official 🤗 Gradio demo</b> for <a href='https://github.com/InstantID/InstantID' target='_blank'><b>InstantID: Zero-shot Identity-Preserving Generation in Seconds</b></a>.<br>
|
||||
|
||||
How to use:<br>
|
||||
1. Upload an image with a face. For images with multiple faces, we will only detect the largest face. Ensure the face is not too small and is clearly visible without significant obstructions or blurring.
|
||||
2. (Optional) You can upload another image as a reference for the face pose. If you don't, we will use the first detected face image to extract facial landmarks. If you use a cropped face at step 1, it is recommended to upload it to define a new face pose.
|
||||
3. (Optional) You can select multiple ControlNet models to control the generation process. The default is to use the IdentityNet only. The ControlNet models include pose skeleton, canny, and depth. You can adjust the strength of each ControlNet model to control the generation process.
|
||||
4. Enter a text prompt, as done in normal text-to-image models.
|
||||
5. Click the <b>Submit</b> button to begin customization.
|
||||
6. Share your customized photo with your friends and enjoy! 😊"""
|
||||
|
||||
article = r"""
|
||||
---
|
||||
📝 **Citation**
|
||||
<br>
|
||||
If our work is helpful for your research or applications, please cite us via:
|
||||
```bibtex
|
||||
@article{wang2024instantid,
|
||||
title={InstantID: Zero-shot Identity-Preserving Generation in Seconds},
|
||||
author={Wang, Qixun and Bai, Xu and Wang, Haofan and Qin, Zekui and Chen, Anthony},
|
||||
journal={arXiv preprint arXiv:2401.07519},
|
||||
year={2024}
|
||||
}
|
||||
```
|
||||
📧 **Contact**
|
||||
<br>
|
||||
If you have any questions, please feel free to open an issue or directly reach us out at <b>haofanwang.ai@gmail.com</b>.
|
||||
"""
|
||||
|
||||
tips = r"""
|
||||
### Usage tips of InstantID
|
||||
1. If you're not satisfied with the similarity, try increasing the weight of "IdentityNet Strength" and "Adapter Strength."
|
||||
2. If you feel that the saturation is too high, first decrease the Adapter strength. If it remains too high, then decrease the IdentityNet strength.
|
||||
3. If you find that text control is not as expected, decrease Adapter strength.
|
||||
4. If you find that realistic style is not good enough, go for our Github repo and use a more realistic base model.
|
||||
"""
|
||||
|
||||
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():
|
||||
with gr.Row(equal_height=True):
|
||||
# upload face image
|
||||
face_file = gr.Image(
|
||||
label="Upload a photo of your face", type="filepath"
|
||||
)
|
||||
# optional: upload a reference pose image
|
||||
pose_file = gr.Image(
|
||||
label="Upload a reference pose image (Optional)",
|
||||
type="filepath",
|
||||
)
|
||||
|
||||
# prompt
|
||||
prompt = gr.Textbox(
|
||||
label="Prompt",
|
||||
info="Give simple prompt is enough to achieve good face fidelity",
|
||||
placeholder="A photo of a person",
|
||||
value="",
|
||||
)
|
||||
|
||||
submit = gr.Button("Submit", variant="primary")
|
||||
enable_LCM = gr.Checkbox(
|
||||
label="Enable Fast Inference with LCM", value=enable_lcm_arg,
|
||||
info="LCM speeds up the inference step, the trade-off is the quality of the generated image. It performs better with portrait face images rather than distant faces",
|
||||
)
|
||||
style = gr.Dropdown(
|
||||
label="Style template",
|
||||
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("Controlnet"):
|
||||
controlnet_selection = gr.CheckboxGroup(
|
||||
["pose", "canny", "depth"], label="Controlnet", value=["pose"],
|
||||
info="Use pose for skeleton inference, canny for edge detection, and depth for depth map estimation. You can try all three to control the generation process"
|
||||
)
|
||||
pose_strength = gr.Slider(
|
||||
label="Pose strength",
|
||||
minimum=0,
|
||||
maximum=1.5,
|
||||
step=0.05,
|
||||
value=0.40,
|
||||
)
|
||||
canny_strength = gr.Slider(
|
||||
label="Canny strength",
|
||||
minimum=0,
|
||||
maximum=1.5,
|
||||
step=0.05,
|
||||
value=0.40,
|
||||
)
|
||||
depth_strength = gr.Slider(
|
||||
label="Depth strength",
|
||||
minimum=0,
|
||||
maximum=1.5,
|
||||
step=0.05,
|
||||
value=0.40,
|
||||
)
|
||||
with gr.Accordion(open=False, label="Advanced Options"):
|
||||
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=1,
|
||||
maximum=100,
|
||||
step=1,
|
||||
value=5 if enable_lcm_arg else 30,
|
||||
)
|
||||
guidance_scale = gr.Slider(
|
||||
label="Guidance scale",
|
||||
minimum=0.1,
|
||||
maximum=20.0,
|
||||
step=0.1,
|
||||
value=0.0 if enable_lcm_arg else 5.0,
|
||||
)
|
||||
seed = gr.Slider(
|
||||
label="Seed",
|
||||
minimum=0,
|
||||
maximum=MAX_SEED,
|
||||
step=1,
|
||||
value=42,
|
||||
)
|
||||
schedulers = [
|
||||
"DEISMultistepScheduler",
|
||||
"HeunDiscreteScheduler",
|
||||
"EulerDiscreteScheduler",
|
||||
"DPMSolverMultistepScheduler",
|
||||
"DPMSolverMultistepScheduler-Karras",
|
||||
"DPMSolverMultistepScheduler-Karras-SDE",
|
||||
]
|
||||
scheduler = gr.Dropdown(
|
||||
label="Schedulers",
|
||||
choices=schedulers,
|
||||
value="EulerDiscreteScheduler",
|
||||
)
|
||||
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
||||
enhance_face_region = gr.Checkbox(label="Enhance non-face region", value=True)
|
||||
|
||||
with gr.Column(scale=1):
|
||||
gallery = gr.Image(label="Generated Images")
|
||||
usage_tips = gr.Markdown(
|
||||
label="InstantID Usage Tips", 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,
|
||||
pose_strength,
|
||||
canny_strength,
|
||||
depth_strength,
|
||||
controlnet_selection,
|
||||
guidance_scale,
|
||||
seed,
|
||||
scheduler,
|
||||
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, pose_file, prompt, style, negative_prompt],
|
||||
fn=run_for_examples,
|
||||
outputs=[gallery, usage_tips],
|
||||
cache_examples=True,
|
||||
)
|
||||
|
||||
gr.Markdown(article)
|
||||
|
||||
demo.launch()
|
||||
|
||||
|
||||
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)
|
||||
|
|
@ -15,6 +15,7 @@ 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
|
||||
|
||||
|
|
@ -22,7 +23,7 @@ import insightface
|
|||
from insightface.app import FaceAnalysis
|
||||
|
||||
from style_template import styles
|
||||
from pipeline_stable_diffusion_xl_instantid import StableDiffusionXLInstantIDPipeline
|
||||
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
|
||||
|
|
@ -45,7 +46,7 @@ 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"):
|
||||
def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8", enable_lcm_arg=False):
|
||||
|
||||
if pretrained_model_name_or_path.endswith(
|
||||
".ckpt"
|
||||
|
|
@ -86,52 +87,57 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
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 swap_to_gallery(images):
|
||||
return gr.update(value=images, visible=True), gr.update(visible=True), gr.update(visible=False)
|
||||
|
||||
def upload_example_to_gallery(images, prompt, style, negative_prompt):
|
||||
return gr.update(value=images, visible=True), gr.update(visible=True), gr.update(visible=False)
|
||||
|
||||
def remove_back_to_files():
|
||||
return gr.update(visible=False), gr.update(visible=False), gr.update(visible=True)
|
||||
|
||||
|
||||
def remove_tips():
|
||||
return gr.update(visible=False)
|
||||
|
||||
def get_example():
|
||||
case = [
|
||||
[
|
||||
['./examples/yann-lecun_resize.jpg'],
|
||||
'./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'],
|
||||
'./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'],
|
||||
'./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'],
|
||||
'./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'],
|
||||
'./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",
|
||||
|
|
@ -139,6 +145,9 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
]
|
||||
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))
|
||||
|
||||
|
|
@ -200,9 +209,15 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
p, n = styles.get(style_name, styles[DEFAULT_STYLE_NAME])
|
||||
return p.replace("{prompt}", positive), n + ' ' + negative
|
||||
|
||||
def generate_image(face_image, pose_image, prompt, negative_prompt, style_name, num_steps, identitynet_strength_ratio, adapter_strength_ratio, guidance_scale, seed, progress=gr.Progress(track_tqdm=True)):
|
||||
|
||||
if face_image is None:
|
||||
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:
|
||||
|
|
@ -211,7 +226,7 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
# apply the style template
|
||||
prompt, negative_prompt = apply_style(style_name, prompt, negative_prompt)
|
||||
|
||||
face_image = load_image(face_image[0])
|
||||
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
|
||||
|
|
@ -226,8 +241,8 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
face_emb = face_info['embedding']
|
||||
face_kps = draw_kps(convert_from_cv2_to_image(face_image_cv2), face_info['kps'])
|
||||
|
||||
if pose_image is not None:
|
||||
pose_image = load_image(pose_image[0])
|
||||
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)
|
||||
|
||||
|
|
@ -240,7 +255,16 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
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...")
|
||||
|
|
@ -252,6 +276,7 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
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,
|
||||
|
|
@ -260,7 +285,7 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
generator=generator
|
||||
).images
|
||||
|
||||
return images, gr.update(visible=True)
|
||||
return images[0], gr.update(visible=True)
|
||||
|
||||
### Description
|
||||
title = r"""
|
||||
|
|
@ -271,11 +296,11 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
<b>Official 🤗 Gradio demo</b> for <a href='https://github.com/InstantID/InstantID' target='_blank'><b>InstantID: Zero-shot Identity-Preserving Generation in Seconds</b></a>.<br>
|
||||
|
||||
How to use:<br>
|
||||
1. Upload a person image. For multiple person images, we will only detect the biggest face. Make sure face is not too small and not significantly blocked or blurred.
|
||||
2. (Optionally) upload another person image as reference pose. If not uploaded, we will use the first person image to extract landmarks. If you use a cropped face at step1, it is recommeneded to upload it to extract a new pose.
|
||||
3. Enter a text prompt as done in normal text-to-image models.
|
||||
4. Click the <b>Submit</b> button to start customizing.
|
||||
5. Share your customizd photo with your friends, enjoy😊!
|
||||
1. Upload an image with a face. For images with multiple faces, we will only detect the largest face. Ensure the face is not too small and is clearly visible without significant obstructions or blurring.
|
||||
2. (Optional) You can upload another image as a reference for the face pose. If you don't, we will use the first detected face image to extract facial landmarks. If you use a cropped face at step 1, it is recommended to upload it to define a new face pose.
|
||||
3. Enter a text prompt, as done in normal text-to-image models.
|
||||
4. Click the <b>Submit</b> button to begin customization.
|
||||
5. Share your customized photo with your friends and enjoy! 😊
|
||||
"""
|
||||
|
||||
article = r"""
|
||||
|
|
@ -298,8 +323,8 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
|
||||
tips = r"""
|
||||
### Usage tips of InstantID
|
||||
1. If you're not satisfied with the similarity, try to increase the weight of "IdentityNet Strength" and "Adapter Strength".
|
||||
2. If you feel that the saturation is too high, first decrease the Adapter strength. If it is still too high, then decrease the IdentityNet strength.
|
||||
1. If you're not satisfied with the similarity, try increasing the weight of "IdentityNet Strength" and "Adapter Strength."
|
||||
2. If you feel that the saturation is too high, first decrease the Adapter strength. If it remains too high, then decrease the IdentityNet strength.
|
||||
3. If you find that text control is not as expected, decrease Adapter strength.
|
||||
4. If you find that realistic style is not good enough, go for our Github repo and use a more realistic base model.
|
||||
"""
|
||||
|
|
@ -317,23 +342,11 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
with gr.Column():
|
||||
|
||||
# upload face image
|
||||
face_files = gr.Files(
|
||||
label="Upload a photo of your face",
|
||||
file_types=["image"]
|
||||
)
|
||||
uploaded_faces = gr.Gallery(label="Your images", visible=False, columns=1, rows=1, height=512)
|
||||
with gr.Column(visible=False) as clear_button_face:
|
||||
remove_and_reupload_faces = gr.ClearButton(value="Remove and upload new ones", components=face_files, size="sm")
|
||||
|
||||
face_file = gr.Image(label="Upload a photo of your face", type="filepath")
|
||||
|
||||
# optional: upload a reference pose image
|
||||
pose_files = gr.Files(
|
||||
label="Upload a reference pose image (optional)",
|
||||
file_types=["image"]
|
||||
)
|
||||
uploaded_poses = gr.Gallery(label="Your images", visible=False, columns=1, rows=1, height=512)
|
||||
with gr.Column(visible=False) as clear_button_pose:
|
||||
remove_and_reupload_poses = gr.ClearButton(value="Remove and upload new ones", components=pose_files, size="sm")
|
||||
|
||||
pose_file = gr.Image(label="Upload a reference pose image (optional)", type="filepath")
|
||||
|
||||
# prompt
|
||||
prompt = gr.Textbox(label="Prompt",
|
||||
info="Give simple prompt is enough to achieve good face fidelity",
|
||||
|
|
@ -342,6 +355,10 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
|
||||
submit = gr.Button("Submit", variant="primary")
|
||||
|
||||
enable_LCM = gr.Checkbox(
|
||||
label="Enable Fast Inference with LCM", value=enable_lcm_arg,
|
||||
info="LCM speeds up the inference step, the trade-off is the quality of the generated image. It performs better with portrait face images rather than distant faces",
|
||||
)
|
||||
style = gr.Dropdown(label="Style template", choices=STYLE_NAMES, value=DEFAULT_STYLE_NAME)
|
||||
|
||||
# strength
|
||||
|
|
@ -371,14 +388,14 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
minimum=20,
|
||||
maximum=100,
|
||||
step=1,
|
||||
value=30,
|
||||
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=5,
|
||||
value=0 if enable_lcm_arg else 5,
|
||||
)
|
||||
seed = gr.Slider(
|
||||
label="Seed",
|
||||
|
|
@ -388,17 +405,12 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
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.Gallery(label="Generated Images")
|
||||
gallery = gr.Image(label="Generated Images")
|
||||
usage_tips = gr.Markdown(label="Usage tips of InstantID", value=tips ,visible=False)
|
||||
|
||||
face_files.upload(fn=swap_to_gallery, inputs=face_files, outputs=[uploaded_faces, clear_button_face, face_files])
|
||||
pose_files.upload(fn=swap_to_gallery, inputs=pose_files, outputs=[uploaded_poses, clear_button_pose, pose_files])
|
||||
|
||||
remove_and_reupload_faces.click(fn=remove_back_to_files, outputs=[uploaded_faces, clear_button_face, face_files])
|
||||
remove_and_reupload_poses.click(fn=remove_back_to_files, outputs=[uploaded_poses, clear_button_pose, pose_files])
|
||||
|
||||
submit.click(
|
||||
fn=remove_tips,
|
||||
outputs=usage_tips,
|
||||
|
|
@ -410,16 +422,19 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
api_name=False,
|
||||
).then(
|
||||
fn=generate_image,
|
||||
inputs=[face_files, pose_files, prompt, negative_prompt, style, num_steps, identitynet_strength_ratio, adapter_strength_ratio, guidance_scale, seed],
|
||||
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_files, prompt, style, negative_prompt],
|
||||
inputs=[face_file, prompt, style, negative_prompt],
|
||||
run_on_click=True,
|
||||
fn=upload_example_to_gallery,
|
||||
outputs=[uploaded_faces, clear_button_face, face_files],
|
||||
fn=run_for_examples,
|
||||
outputs=[gallery, usage_tips],
|
||||
cache_examples=True,
|
||||
)
|
||||
|
||||
gr.Markdown(article)
|
||||
|
|
@ -428,9 +443,9 @@ def main(pretrained_model_name_or_path="wangqixun/YamerMIX_v8"):
|
|||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument(
|
||||
"--pretrained_model_name_or_path", type=str, default="wangqixun/YamerMIX_v8"
|
||||
)
|
||||
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)
|
||||
main(args.pretrained_model_name_or_path, args.enable_LCM)
|
||||
|
|
@ -0,0 +1,39 @@
|
|||
import torch
|
||||
import numpy as np
|
||||
from PIL import Image
|
||||
from controlnet_aux import OpenposeDetector
|
||||
from model_util import get_torch_device
|
||||
import cv2
|
||||
|
||||
|
||||
from transformers import DPTImageProcessor, DPTForDepthEstimation
|
||||
|
||||
device = get_torch_device()
|
||||
depth_estimator = DPTForDepthEstimation.from_pretrained("Intel/dpt-hybrid-midas").to(device)
|
||||
feature_extractor = DPTImageProcessor.from_pretrained("Intel/dpt-hybrid-midas")
|
||||
openpose = OpenposeDetector.from_pretrained("lllyasviel/ControlNet")
|
||||
|
||||
def get_depth_map(image):
|
||||
image = feature_extractor(images=image, return_tensors="pt").pixel_values.to("cuda")
|
||||
with torch.no_grad(), torch.autocast("cuda"):
|
||||
depth_map = depth_estimator(image).predicted_depth
|
||||
|
||||
depth_map = torch.nn.functional.interpolate(
|
||||
depth_map.unsqueeze(1),
|
||||
size=(1024, 1024),
|
||||
mode="bicubic",
|
||||
align_corners=False,
|
||||
)
|
||||
depth_min = torch.amin(depth_map, dim=[1, 2, 3], keepdim=True)
|
||||
depth_max = torch.amax(depth_map, dim=[1, 2, 3], keepdim=True)
|
||||
depth_map = (depth_map - depth_min) / (depth_max - depth_min)
|
||||
image = torch.cat([depth_map] * 3, dim=1)
|
||||
|
||||
image = image.permute(0, 2, 3, 1).cpu().numpy()[0]
|
||||
image = Image.fromarray((image * 255.0).clip(0, 255).astype(np.uint8))
|
||||
return image
|
||||
|
||||
def get_canny_image(image, t1=100, t2=200):
|
||||
image = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
|
||||
edges = cv2.Canny(image, t1, t2)
|
||||
return Image.fromarray(edges, "L")
|
||||
|
|
@ -0,0 +1,27 @@
|
|||
from huggingface_hub import hf_hub_download
|
||||
import gdown
|
||||
import os
|
||||
|
||||
# download models
|
||||
hf_hub_download(
|
||||
repo_id="InstantX/InstantID",
|
||||
filename="ControlNetModel/config.json",
|
||||
local_dir="./checkpoints",
|
||||
)
|
||||
hf_hub_download(
|
||||
repo_id="InstantX/InstantID",
|
||||
filename="ControlNetModel/diffusion_pytorch_model.safetensors",
|
||||
local_dir="./checkpoints",
|
||||
)
|
||||
hf_hub_download(
|
||||
repo_id="InstantX/InstantID", filename="ip-adapter.bin", local_dir="./checkpoints"
|
||||
)
|
||||
hf_hub_download(
|
||||
repo_id="latent-consistency/lcm-lora-sdxl",
|
||||
filename="pytorch_lora_weights.safetensors",
|
||||
local_dir="./checkpoints",
|
||||
)
|
||||
# download antelopev2
|
||||
gdown.download(url="https://drive.google.com/file/d/18wEUfMNohBJ4K3Ly5wpTejPfDzp-8fI8/view?usp=sharing", output="./models/", quiet=False, fuzzy=True)
|
||||
# unzip antelopev2.zip
|
||||
os.system("unzip ./models/antelopev2.zip -d ./models/")
|
||||
|
|
@ -1,7 +1,7 @@
|
|||
diffusers==0.25.0
|
||||
diffusers==0.25.1
|
||||
torch==2.0.0
|
||||
torchvision==0.15.1
|
||||
transformers==4.30.1
|
||||
transformers==4.37.1
|
||||
accelerate
|
||||
safetensors
|
||||
einops
|
||||
|
|
@ -12,4 +12,7 @@ peft
|
|||
huggingface-hub==0.20.2
|
||||
opencv-python
|
||||
insightface
|
||||
gradio
|
||||
gradio
|
||||
controlnet_aux
|
||||
gdown
|
||||
peft
|
||||
2
infer.py
2
infer.py
|
|
@ -35,6 +35,7 @@ def resize_img(input_image, max_side=1280, min_side=1024, size=None,
|
|||
|
||||
if __name__ == "__main__":
|
||||
|
||||
# Load face encoder
|
||||
app = FaceAnalysis(name='antelopev2', root='./', providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
|
||||
app.prepare(ctx_id=0, det_size=(640, 640))
|
||||
|
||||
|
|
@ -55,6 +56,7 @@ if __name__ == "__main__":
|
|||
pipe.cuda()
|
||||
pipe.load_ip_adapter_instantid(face_adapter)
|
||||
|
||||
# Infer setting
|
||||
prompt = "analog film photo of a man. faded film, desaturated, 35mm photo, grainy, vignette, vintage, Kodachrome, Lomography, stained, highly detailed, found footage, masterpiece, best quality"
|
||||
n_prompt = "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, painting, drawing, illustration, glitch, deformed, mutated, cross-eyed, ugly, disfigured (lowres, low quality, worst quality:1.2), (text:1.2), watermark, painting, drawing, illustration, glitch,deformed, mutated, cross-eyed, ugly, disfigured"
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,119 @@
|
|||
import cv2
|
||||
import torch
|
||||
import numpy as np
|
||||
from PIL import Image
|
||||
|
||||
from diffusers.utils import load_image
|
||||
from diffusers.models import ControlNetModel
|
||||
from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
|
||||
|
||||
from insightface.app import FaceAnalysis
|
||||
from pipeline_stable_diffusion_xl_instantid_full import StableDiffusionXLInstantIDPipeline, draw_kps
|
||||
|
||||
from controlnet_aux import MidasDetector
|
||||
|
||||
def convert_from_image_to_cv2(img: Image) -> np.ndarray:
|
||||
return cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
|
||||
|
||||
def resize_img(input_image, max_side=1280, min_side=1024, size=None,
|
||||
pad_to_max_side=False, mode=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
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
# 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'
|
||||
controlnet_depth_path = f'diffusers/controlnet-depth-sdxl-1.0-small'
|
||||
|
||||
# Load depth detector
|
||||
midas = MidasDetector.from_pretrained("lllyasviel/Annotators")
|
||||
|
||||
# Load pipeline
|
||||
controlnet_list = [controlnet_path, controlnet_depth_path]
|
||||
controlnet_model_list = []
|
||||
for controlnet_path in controlnet_list:
|
||||
controlnet = ControlNetModel.from_pretrained(controlnet_path, torch_dtype=torch.float16)
|
||||
controlnet_model_list.append(controlnet)
|
||||
controlnet = MultiControlNetModel(controlnet_model_list)
|
||||
|
||||
base_model_path = 'stabilityai/stable-diffusion-xl-base-1.0'
|
||||
|
||||
pipe = StableDiffusionXLInstantIDPipeline.from_pretrained(
|
||||
base_model_path,
|
||||
controlnet=controlnet,
|
||||
torch_dtype=torch.float16,
|
||||
)
|
||||
pipe.cuda()
|
||||
pipe.load_ip_adapter_instantid(face_adapter)
|
||||
|
||||
# Infer setting
|
||||
prompt = "analog film photo of a man. faded film, desaturated, 35mm photo, grainy, vignette, vintage, Kodachrome, Lomography, stained, highly detailed, found footage, masterpiece, best quality"
|
||||
n_prompt = "(lowres, low quality, worst quality:1.2), (text:1.2), watermark, painting, drawing, illustration, glitch, deformed, mutated, cross-eyed, ugly, disfigured (lowres, low quality, worst quality:1.2), (text:1.2), watermark, painting, drawing, illustration, glitch,deformed, mutated, cross-eyed, ugly, disfigured"
|
||||
|
||||
face_image = load_image("./examples/yann-lecun_resize.jpg")
|
||||
face_image = resize_img(face_image)
|
||||
|
||||
face_info = app.get(cv2.cvtColor(np.array(face_image), cv2.COLOR_RGB2BGR))
|
||||
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']
|
||||
|
||||
# use another reference image
|
||||
pose_image = load_image("./examples/poses/pose.jpg")
|
||||
pose_image = resize_img(pose_image)
|
||||
|
||||
face_info = app.get(cv2.cvtColor(np.array(pose_image), cv2.COLOR_RGB2BGR))
|
||||
pose_image_cv2 = convert_from_image_to_cv2(pose_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_kps = draw_kps(pose_image, face_info['kps'])
|
||||
|
||||
width, height = face_kps.size
|
||||
|
||||
# use depth control
|
||||
processed_image_midas = midas(pose_image)
|
||||
processed_image_midas = processed_image_midas.resize(pose_image.size)
|
||||
|
||||
# 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))
|
||||
|
||||
image = pipe(
|
||||
prompt=prompt,
|
||||
negative_prompt=n_prompt,
|
||||
image_embeds=face_emb,
|
||||
control_mask=control_mask,
|
||||
image=[face_kps, processed_image_midas],
|
||||
controlnet_conditioning_scale=[0.8,0.8],
|
||||
ip_adapter_scale=0.8,
|
||||
num_inference_steps=30,
|
||||
guidance_scale=5,
|
||||
).images[0]
|
||||
|
||||
image.save('result.jpg')
|
||||
|
|
@ -10,6 +10,10 @@ try:
|
|||
except Exception as e:
|
||||
xformers_available = False
|
||||
|
||||
class RegionControler(object):
|
||||
def __init__(self) -> None:
|
||||
self.prompt_image_conditioning = []
|
||||
region_control = RegionControler()
|
||||
|
||||
class AttnProcessor(nn.Module):
|
||||
r"""
|
||||
|
|
@ -22,7 +26,7 @@ class AttnProcessor(nn.Module):
|
|||
):
|
||||
super().__init__()
|
||||
|
||||
def __call__(
|
||||
def forward(
|
||||
self,
|
||||
attn,
|
||||
hidden_states,
|
||||
|
|
@ -108,7 +112,7 @@ class IPAttnProcessor(nn.Module):
|
|||
self.to_k_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
|
||||
self.to_v_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
|
||||
|
||||
def __call__(
|
||||
def forward(
|
||||
self,
|
||||
attn,
|
||||
hidden_states,
|
||||
|
|
@ -174,6 +178,17 @@ class IPAttnProcessor(nn.Module):
|
|||
ip_hidden_states = torch.bmm(ip_attention_probs, ip_value)
|
||||
ip_hidden_states = attn.batch_to_head_dim(ip_hidden_states)
|
||||
|
||||
# region control
|
||||
if len(region_control.prompt_image_conditioning) == 1:
|
||||
region_mask = region_control.prompt_image_conditioning[0].get('region_mask', None)
|
||||
if region_mask is not None:
|
||||
h, w = region_mask.shape[:2]
|
||||
ratio = (h * w / query.shape[1]) ** 0.5
|
||||
mask = F.interpolate(region_mask[None, None], scale_factor=1/ratio, mode='nearest').reshape([1, -1, 1])
|
||||
else:
|
||||
mask = torch.ones_like(ip_hidden_states)
|
||||
ip_hidden_states = ip_hidden_states * mask
|
||||
|
||||
hidden_states = hidden_states + self.scale * ip_hidden_states
|
||||
|
||||
# linear proj
|
||||
|
|
@ -215,7 +230,7 @@ class AttnProcessor2_0(torch.nn.Module):
|
|||
if not hasattr(F, "scaled_dot_product_attention"):
|
||||
raise ImportError("AttnProcessor2_0 requires PyTorch 2.0, to use it, please upgrade PyTorch to 2.0.")
|
||||
|
||||
def __call__(
|
||||
def forward(
|
||||
self,
|
||||
attn,
|
||||
hidden_states,
|
||||
|
|
@ -317,7 +332,7 @@ class IPAttnProcessor2_0(torch.nn.Module):
|
|||
self.to_k_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
|
||||
self.to_v_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
|
||||
|
||||
def __call__(
|
||||
def forward(
|
||||
self,
|
||||
attn,
|
||||
hidden_states,
|
||||
|
|
@ -402,6 +417,17 @@ class IPAttnProcessor2_0(torch.nn.Module):
|
|||
ip_hidden_states = ip_hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
|
||||
ip_hidden_states = ip_hidden_states.to(query.dtype)
|
||||
|
||||
# region control
|
||||
if len(region_control.prompt_image_conditioning) == 1:
|
||||
region_mask = region_control.prompt_image_conditioning[0].get('region_mask', None)
|
||||
if region_mask is not None:
|
||||
h, w = region_mask.shape[:2]
|
||||
ratio = (h * w / query.shape[1]) ** 0.5
|
||||
mask = F.interpolate(region_mask[None, None], scale_factor=1/ratio, mode='nearest').reshape([1, -1, 1])
|
||||
else:
|
||||
mask = torch.ones_like(ip_hidden_states)
|
||||
ip_hidden_states = ip_hidden_states * mask
|
||||
|
||||
hidden_states = hidden_states + self.scale * ip_hidden_states
|
||||
|
||||
# linear proj
|
||||
|
|
|
|||
|
|
@ -767,4 +767,4 @@ class StableDiffusionXLInstantIDPipeline(StableDiffusionXLControlNetPipeline):
|
|||
if not return_dict:
|
||||
return (image,)
|
||||
|
||||
return StableDiffusionXLPipelineOutput(images=image)
|
||||
return StableDiffusionXLPipelineOutput(images=image)
|
||||
File diff suppressed because it is too large
Load Diff
Loading…
Reference in New Issue