Spaces:
Runtime error
Runtime error
add threshold and negative prompt
Browse files
app.py
CHANGED
@@ -10,10 +10,10 @@ import cv2
|
|
10 |
import torch
|
11 |
|
12 |
from diffusers import StableDiffusion3ControlNetPipeline
|
13 |
-
from diffusers.models import SD3ControlNetModel
|
14 |
from diffusers.utils import load_image
|
15 |
|
16 |
-
#
|
17 |
controlnet_canny = SD3ControlNetModel.from_pretrained("InstantX/SD3-Controlnet-Canny")
|
18 |
pipe = StableDiffusion3ControlNetPipeline.from_pretrained(
|
19 |
"stabilityai/stable-diffusion-3-medium-diffusers",
|
@@ -41,38 +41,35 @@ def resize_image(input_path, output_path, target_height):
|
|
41 |
|
42 |
|
43 |
@spaces.GPU(duration=90)
|
44 |
-
def infer(image_in, prompt, inference_steps, guidance_scale, control_weight, progress=gr.Progress(track_tqdm=True)):
|
45 |
-
|
46 |
-
n_prompt = 'NSFW, nude, naked, porn, ugly'
|
47 |
|
48 |
# Canny preprocessing
|
49 |
image_to_canny = load_image(image_in)
|
50 |
image_to_canny = np.array(image_to_canny)
|
51 |
-
image_to_canny = cv2.Canny(image_to_canny,
|
52 |
image_to_canny = image_to_canny[:, :, None]
|
53 |
image_to_canny = np.concatenate([image_to_canny, image_to_canny, image_to_canny], axis=2)
|
54 |
image_to_canny = Image.fromarray(image_to_canny)
|
55 |
|
56 |
control_image = image_to_canny
|
57 |
|
58 |
-
#
|
59 |
image = pipe(
|
60 |
prompt=prompt,
|
61 |
-
negative_prompt=
|
62 |
control_image=control_image,
|
63 |
controlnet_conditioning_scale=control_weight,
|
64 |
num_inference_steps=inference_steps,
|
65 |
guidance_scale=guidance_scale,
|
66 |
).images[0]
|
67 |
|
68 |
-
|
69 |
image_redim, w, h = resize_image(image_in, "resized_input.jpg", 1024)
|
70 |
image = image.resize((w, h), Image.LANCZOS)
|
71 |
|
72 |
return image, gr.update(value=image_to_canny, visible=True)
|
73 |
-
|
74 |
|
75 |
-
|
|
|
76 |
#col-container{
|
77 |
margin: 0 auto;
|
78 |
max-width: 1080px;
|
@@ -92,6 +89,7 @@ with gr.Blocks(css=css) as demo:
|
|
92 |
with gr.Column():
|
93 |
image_in = gr.Image(label="Image reference", sources=["upload"], type="filepath")
|
94 |
prompt = gr.Textbox(label="Prompt")
|
|
|
95 |
|
96 |
with gr.Accordion("Advanced settings", open=False):
|
97 |
with gr.Column():
|
@@ -99,22 +97,23 @@ with gr.Blocks(css=css) as demo:
|
|
99 |
inference_steps = gr.Slider(label="Inference steps", minimum=1, maximum=50, step=1, value=25)
|
100 |
guidance_scale = gr.Slider(label="Guidance scale", minimum=1.0, maximum=10.0, step=0.1, value=7.0)
|
101 |
control_weight = gr.Slider(label="Control Weight", minimum=0.0, maximum=1.0, step=0.01, value=0.7)
|
|
|
|
|
102 |
|
103 |
submit_canny_btn = gr.Button("Submit")
|
104 |
|
105 |
with gr.Column():
|
106 |
result = gr.Image(label="Result")
|
107 |
canny_used = gr.Image(label="Preprocessed Canny", visible=False)
|
108 |
-
|
109 |
|
110 |
|
111 |
submit_canny_btn.click(
|
112 |
fn=infer,
|
113 |
-
inputs=[image_in, prompt, inference_steps, guidance_scale, control_weight],
|
114 |
outputs=[result, canny_used],
|
115 |
api_name="predict",
|
116 |
show_api=True
|
117 |
)
|
118 |
|
119 |
-
# Enable API by setting enable_api=True
|
120 |
demo.queue().launch(show_api=True)
|
|
|
|
10 |
import torch
|
11 |
|
12 |
from diffusers import StableDiffusion3ControlNetPipeline
|
13 |
+
from diffusers.models import SD3ControlNetModel
|
14 |
from diffusers.utils import load_image
|
15 |
|
16 |
+
# Load pipeline
|
17 |
controlnet_canny = SD3ControlNetModel.from_pretrained("InstantX/SD3-Controlnet-Canny")
|
18 |
pipe = StableDiffusion3ControlNetPipeline.from_pretrained(
|
19 |
"stabilityai/stable-diffusion-3-medium-diffusers",
|
|
|
41 |
|
42 |
|
43 |
@spaces.GPU(duration=90)
|
44 |
+
def infer(image_in, prompt, negative_prompt, inference_steps, guidance_scale, control_weight, low_threshold, high_threshold, progress=gr.Progress(track_tqdm=True)):
|
|
|
|
|
45 |
|
46 |
# Canny preprocessing
|
47 |
image_to_canny = load_image(image_in)
|
48 |
image_to_canny = np.array(image_to_canny)
|
49 |
+
image_to_canny = cv2.Canny(image_to_canny, low_threshold, high_threshold)
|
50 |
image_to_canny = image_to_canny[:, :, None]
|
51 |
image_to_canny = np.concatenate([image_to_canny, image_to_canny, image_to_canny], axis=2)
|
52 |
image_to_canny = Image.fromarray(image_to_canny)
|
53 |
|
54 |
control_image = image_to_canny
|
55 |
|
56 |
+
# Infer
|
57 |
image = pipe(
|
58 |
prompt=prompt,
|
59 |
+
negative_prompt=negative_prompt,
|
60 |
control_image=control_image,
|
61 |
controlnet_conditioning_scale=control_weight,
|
62 |
num_inference_steps=inference_steps,
|
63 |
guidance_scale=guidance_scale,
|
64 |
).images[0]
|
65 |
|
|
|
66 |
image_redim, w, h = resize_image(image_in, "resized_input.jpg", 1024)
|
67 |
image = image.resize((w, h), Image.LANCZOS)
|
68 |
|
69 |
return image, gr.update(value=image_to_canny, visible=True)
|
|
|
70 |
|
71 |
+
|
72 |
+
css = """
|
73 |
#col-container{
|
74 |
margin: 0 auto;
|
75 |
max-width: 1080px;
|
|
|
89 |
with gr.Column():
|
90 |
image_in = gr.Image(label="Image reference", sources=["upload"], type="filepath")
|
91 |
prompt = gr.Textbox(label="Prompt")
|
92 |
+
negative_prompt = gr.Textbox(label="Negative Prompt", placeholder="Enter negative prompts here")
|
93 |
|
94 |
with gr.Accordion("Advanced settings", open=False):
|
95 |
with gr.Column():
|
|
|
97 |
inference_steps = gr.Slider(label="Inference steps", minimum=1, maximum=50, step=1, value=25)
|
98 |
guidance_scale = gr.Slider(label="Guidance scale", minimum=1.0, maximum=10.0, step=0.1, value=7.0)
|
99 |
control_weight = gr.Slider(label="Control Weight", minimum=0.0, maximum=1.0, step=0.01, value=0.7)
|
100 |
+
low_threshold = gr.Slider(label="Canny Low Threshold", minimum=0, maximum=255, step=1, value=100)
|
101 |
+
high_threshold = gr.Slider(label="Canny High Threshold", minimum=0, maximum=255, step=1, value=200)
|
102 |
|
103 |
submit_canny_btn = gr.Button("Submit")
|
104 |
|
105 |
with gr.Column():
|
106 |
result = gr.Image(label="Result")
|
107 |
canny_used = gr.Image(label="Preprocessed Canny", visible=False)
|
|
|
108 |
|
109 |
|
110 |
submit_canny_btn.click(
|
111 |
fn=infer,
|
112 |
+
inputs=[image_in, prompt, negative_prompt, inference_steps, guidance_scale, control_weight, low_threshold, high_threshold],
|
113 |
outputs=[result, canny_used],
|
114 |
api_name="predict",
|
115 |
show_api=True
|
116 |
)
|
117 |
|
|
|
118 |
demo.queue().launch(show_api=True)
|
119 |
+
|