DiffuseCraft / image_processor.py
r3gm's picture
Create image_processor.py
c2c3c5c verified
import spaces
import gradio as gr
from stablepy import Preprocessor
PREPROCESSOR_TASKS_LIST = [
"Canny",
"Openpose",
"DPT",
"Midas",
"ZoeDepth",
"DepthAnything",
"HED",
"PidiNet",
"TEED",
"Lineart",
"LineartAnime",
"Anyline",
"Lineart standard",
"SegFormer",
"UPerNet",
"ContentShuffle",
"Recolor",
"Blur",
"MLSD",
"NormalBae",
]
preprocessor = Preprocessor()
def process_inputs(
image,
name,
resolution,
precessor_resolution,
low_threshold,
high_threshold,
value_threshod,
distance_threshold,
recolor_mode,
recolor_gamma_correction,
blur_k_size,
pre_openpose_extra,
hed_scribble,
pre_pidinet_safe,
pre_lineart_coarse,
use_cuda,
):
if not image:
raise ValueError("To use this, simply upload an image.")
preprocessor.load(name, False)
params = dict(
image_resolution=resolution,
detect_resolution=precessor_resolution,
low_threshold=low_threshold,
high_threshold=high_threshold,
thr_v=value_threshod,
thr_d=distance_threshold,
mode=recolor_mode,
gamma_correction=recolor_gamma_correction,
blur_sigma=blur_k_size,
hand_and_face=pre_openpose_extra,
scribble=hed_scribble,
safe=pre_pidinet_safe,
coarse=pre_lineart_coarse,
)
if use_cuda:
@spaces.GPU(duration=15)
def wrapped_func():
preprocessor.to("cuda")
return preprocessor(image, **params)
return wrapped_func()
return preprocessor(image, **params)
def preprocessor_tab():
with gr.Row():
with gr.Column():
pre_image = gr.Image(label="Image", type="pil", sources=["upload"])
pre_options = gr.Dropdown(label="Preprocessor", choices=PREPROCESSOR_TASKS_LIST, value=PREPROCESSOR_TASKS_LIST[0])
pre_img_resolution = gr.Slider(
minimum=64, maximum=4096, step=64, value=1024, label="Image Resolution",
info="The maximum proportional size of the generated image based on the uploaded image."
)
pre_start = gr.Button(value="PROCESS IMAGE", variant="primary")
with gr.Accordion("Advanced Settings", open=False):
with gr.Column():
pre_processor_resolution = gr.Slider(minimum=64, maximum=2048, step=64, value=512, label="Preprocessor Resolution")
pre_low_threshold = gr.Slider(minimum=1, maximum=255, step=1, value=100, label="'CANNY' low threshold")
pre_high_threshold = gr.Slider(minimum=1, maximum=255, step=1, value=200, label="'CANNY' high threshold")
pre_value_threshold = gr.Slider(minimum=1, maximum=2.0, step=0.01, value=0.1, label="'MLSD' Hough value threshold")
pre_distance_threshold = gr.Slider(minimum=1, maximum=20.0, step=0.01, value=0.1, label="'MLSD' Hough distance threshold")
pre_recolor_mode = gr.Dropdown(label="'RECOLOR' mode", choices=["luminance", "intensity"], value="luminance")
pre_recolor_gamma_correction = gr.Number(minimum=0., maximum=25., value=1., step=0.001, label="'RECOLOR' gamma correction")
pre_blur_k_size = gr.Number(minimum=0, maximum=100, value=9, step=1, label="'BLUR' sigma")
pre_openpose_extra = gr.Checkbox(value=True, label="'OPENPOSE' face and hand")
pre_hed_scribble = gr.Checkbox(value=False, label="'HED' scribble")
pre_pidinet_safe = gr.Checkbox(value=False, label="'PIDINET' safe")
pre_lineart_coarse = gr.Checkbox(value=False, label="'LINEART' coarse")
pre_use_cuda = gr.Checkbox(value=False, label="Use CUDA")
with gr.Column():
pre_result = gr.Image(label="Result", type="pil", interactive=False, format="png")
pre_start.click(
fn=process_inputs,
inputs=[
pre_image,
pre_options,
pre_img_resolution,
pre_processor_resolution,
pre_low_threshold,
pre_high_threshold,
pre_value_threshold,
pre_distance_threshold,
pre_recolor_mode,
pre_recolor_gamma_correction,
pre_blur_k_size,
pre_openpose_extra,
pre_hed_scribble,
pre_pidinet_safe,
pre_lineart_coarse,
pre_use_cuda,
],
outputs=[pre_result],
)