import gradio as gr
import numpy as np
import torch
from diffusers import ControlNetModel, StableDiffusionControlNetInpaintPipeline
from PIL import Image
from diffusion_webui.diffusion_models.base_controlnet_pipeline import (
ControlnetPipeline,
)
from diffusion_webui.utils.model_list import (
controlnet_model_list,
stable_model_list,
)
from diffusion_webui.utils.preprocces_utils import PREPROCCES_DICT
from diffusion_webui.utils.scheduler_list import (
SCHEDULER_MAPPING,
get_scheduler,
)
class StableDiffusionControlNetInpaintGenerator(ControlnetPipeline):
def __init__(self):
super().__init__()
def load_model(self, stable_model_path, controlnet_model_path, scheduler):
if self.pipe is None or self.pipe.model_name != stable_model_path or self.pipe.scheduler_name != scheduler:
controlnet = ControlNetModel.from_pretrained(
controlnet_model_path, torch_dtype=torch.float16
)
self.pipe = (
StableDiffusionControlNetInpaintPipeline.from_pretrained(
pretrained_model_name_or_path=stable_model_path,
controlnet=controlnet,
safety_checker=None,
torch_dtype=torch.float16,
)
)
self.pipe.model_name = stable_model_path
self.pipe.scheduler_name = scheduler
self.pipe = get_scheduler(pipe=self.pipe, scheduler=scheduler)
self.pipe.to("cuda")
self.pipe.enable_xformers_memory_efficient_attention()
return self.pipe
def load_image(self, image):
image = np.array(image)
image = Image.fromarray(image)
return image
def controlnet_preprocces(
self,
read_image: str,
preprocces_type: str,
):
processed_image = PREPROCCES_DICT[preprocces_type](read_image)
return processed_image
def generate_image(
self,
image_path: str,
stable_model_path: str,
controlnet_model_path: str,
prompt: str,
negative_prompt: str,
num_images_per_prompt: int,
height: int,
width: int,
strength: int,
guess_mode: bool,
guidance_scale: int,
num_inference_step: int,
controlnet_conditioning_scale: int,
scheduler: str,
seed_generator: int,
preprocces_type: str,
):
normal_image = image_path["image"].convert("RGB").resize((512, 512))
mask_image = image_path["mask"].convert("RGB").resize((512, 512))
normal_image = self.load_image(image=normal_image)
mask_image = self.load_image(image=mask_image)
control_image = self.controlnet_preprocces(
read_image=normal_image, preprocces_type=preprocces_type
)
pipe = self.load_model(
stable_model_path=stable_model_path,
controlnet_model_path=controlnet_model_path,
scheduler=scheduler,
)
if seed_generator == 0:
random_seed = torch.randint(0, 1000000, (1,))
generator = torch.manual_seed(random_seed)
else:
generator = torch.manual_seed(seed_generator)
output = pipe(
prompt=prompt,
image=normal_image,
height=height,
width=width,
mask_image=mask_image,
strength=strength,
guess_mode=guess_mode,
control_image=control_image,
negative_prompt=negative_prompt,
num_images_per_prompt=num_images_per_prompt,
num_inference_steps=num_inference_step,
guidance_scale=guidance_scale,
controlnet_conditioning_scale=float(controlnet_conditioning_scale),
generator=generator,
).images
return output
def app():
with gr.Blocks():
with gr.Row():
with gr.Column():
controlnet_inpaint_image_path = gr.Image(
source="upload",
tool="sketch",
elem_id="image_upload",
type="pil",
label="Upload",
).style(height=260)
controlnet_inpaint_prompt = gr.Textbox(
lines=1, placeholder="Prompt", show_label=False
)
controlnet_inpaint_negative_prompt = gr.Textbox(
lines=1, placeholder="Negative Prompt", show_label=False
)
with gr.Row():
with gr.Column():
controlnet_inpaint_stable_model_path = gr.Dropdown(
choices=stable_model_list,
value=stable_model_list[0],
label="Stable Model Path",
)
controlnet_inpaint_preprocces_type = gr.Dropdown(
choices=list(PREPROCCES_DICT.keys()),
value=list(PREPROCCES_DICT.keys())[0],
label="Preprocess Type",
)
controlnet_inpaint_conditioning_scale = gr.Slider(
minimum=0.0,
maximum=1.0,
step=0.1,
value=1.0,
label="ControlNet Conditioning Scale",
)
controlnet_inpaint_guidance_scale = gr.Slider(
minimum=0.1,
maximum=15,
step=0.1,
value=7.5,
label="Guidance Scale",
)
controlnet_inpaint_height = gr.Slider(
minimum=128,
maximum=1280,
step=32,
value=512,
label="Height",
)
controlnet_inpaint_width = gr.Slider(
minimum=128,
maximum=1280,
step=32,
value=512,
label="Width",
)
controlnet_inpaint_guess_mode = gr.Checkbox(
label="Guess Mode"
)
with gr.Column():
controlnet_inpaint_model_path = gr.Dropdown(
choices=controlnet_model_list,
value=controlnet_model_list[0],
label="ControlNet Model Path",
)
controlnet_inpaint_scheduler = gr.Dropdown(
choices=list(SCHEDULER_MAPPING.keys()),
value=list(SCHEDULER_MAPPING.keys())[0],
label="Scheduler",
)
controlnet_inpaint_strength = gr.Slider(
minimum=0.1,
maximum=15,
step=0.1,
value=7.5,
label="Strength",
)
controlnet_inpaint_num_inference_step = gr.Slider(
minimum=1,
maximum=150,
step=1,
value=30,
label="Num Inference Step",
)
controlnet_inpaint_num_images_per_prompt = (
gr.Slider(
minimum=1,
maximum=4,
step=1,
value=1,
label="Number Of Images",
)
)
controlnet_inpaint_seed_generator = gr.Slider(
minimum=0,
maximum=1000000,
step=1,
value=0,
label="Seed(0 for random)",
)
# Button to generate the image
controlnet_inpaint_predict_button = gr.Button(
value="Generate Image"
)
with gr.Column():
# Gallery to display the generated images
controlnet_inpaint_output_image = gr.Gallery(
label="Generated images",
show_label=False,
elem_id="gallery",
).style(grid=(1, 2))
controlnet_inpaint_predict_button.click(
fn=StableDiffusionControlNetInpaintGenerator().generate_image,
inputs=[
controlnet_inpaint_image_path,
controlnet_inpaint_stable_model_path,
controlnet_inpaint_model_path,
controlnet_inpaint_prompt,
controlnet_inpaint_negative_prompt,
controlnet_inpaint_num_images_per_prompt,
controlnet_inpaint_height,
controlnet_inpaint_width,
controlnet_inpaint_strength,
controlnet_inpaint_guess_mode,
controlnet_inpaint_guidance_scale,
controlnet_inpaint_num_inference_step,
controlnet_inpaint_conditioning_scale,
controlnet_inpaint_scheduler,
controlnet_inpaint_seed_generator,
controlnet_inpaint_preprocces_type,
],
outputs=[controlnet_inpaint_output_image],
)