Spaces:
Runtime error
Runtime error
Commit
·
0c8abab
1
Parent(s):
5d45a2a
Upload QRDiffuser.py
Browse files- QRDiffuser.py +227 -0
QRDiffuser.py
ADDED
@@ -0,0 +1,227 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import gradio as gr
|
3 |
+
from PIL import Image
|
4 |
+
import qrcode
|
5 |
+
import os
|
6 |
+
|
7 |
+
from diffusers import (
|
8 |
+
StableDiffusionControlNetPipeline,
|
9 |
+
ControlNetModel,
|
10 |
+
DDIMScheduler,
|
11 |
+
DPMSolverMultistepScheduler,
|
12 |
+
UniPCMultistepScheduler,
|
13 |
+
DEISMultistepScheduler,
|
14 |
+
HeunDiscreteScheduler,
|
15 |
+
EulerDiscreteScheduler,
|
16 |
+
EulerAncestralDiscreteScheduler,
|
17 |
+
)
|
18 |
+
|
19 |
+
controlnet = ControlNetModel.from_pretrained(
|
20 |
+
"monster-labs/control_v1p_sd15_qrcode_monster",
|
21 |
+
torch_dtype=torch.float16,
|
22 |
+
)
|
23 |
+
|
24 |
+
pipe = StableDiffusionControlNetPipeline.from_pretrained(
|
25 |
+
#"runwayml/stable-diffusion-v1-5",
|
26 |
+
"SG161222/Realistic_Vision_V3.0_VAE",
|
27 |
+
controlnet=controlnet,
|
28 |
+
safety_checker=None,
|
29 |
+
torch_dtype=torch.float16,
|
30 |
+
).to("cuda")
|
31 |
+
#pipe.enable_xformers_memory_efficient_attention()
|
32 |
+
pipe.enable_attention_slicing(1)
|
33 |
+
pipe.enable_model_cpu_offload()
|
34 |
+
#pipe.enable_vae_tiling()
|
35 |
+
pipe.enable_vae_slicing()
|
36 |
+
pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
|
37 |
+
|
38 |
+
SAMPLER_MAP = {
|
39 |
+
"DPM++ Karras SDE": lambda config: DPMSolverMultistepScheduler.from_config(config, use_karras=True, algorithm_type="sde-dpmsolver++"),
|
40 |
+
"DPM++ Karras": lambda config: DPMSolverMultistepScheduler.from_config(config, use_karras=True),
|
41 |
+
"Heun": lambda config: HeunDiscreteScheduler.from_config(config),
|
42 |
+
"Euler a": lambda config: EulerAncestralDiscreteScheduler.from_config(config),
|
43 |
+
"Euler": lambda config: EulerDiscreteScheduler.from_config(config),
|
44 |
+
"DDIM": lambda config: DDIMScheduler.from_config(config),
|
45 |
+
"DEIS": lambda config: DEISMultistepScheduler.from_config(config),
|
46 |
+
}
|
47 |
+
|
48 |
+
boxsize=16
|
49 |
+
def create_code(content: str, errorCorrection: str):
|
50 |
+
match errorCorrection:
|
51 |
+
case "L 7%":
|
52 |
+
errCorr = qrcode.constants.ERROR_CORRECT_L
|
53 |
+
case "M 15%":
|
54 |
+
errCorr = qrcode.constants.ERROR_CORRECT_M
|
55 |
+
case "Q 25%":
|
56 |
+
errCorr = qrcode.constants.ERROR_CORRECT_Q
|
57 |
+
case "H 30%":
|
58 |
+
errCorr = qrcode.constants.ERROR_CORRECT_H
|
59 |
+
|
60 |
+
qr = qrcode.QRCode(
|
61 |
+
version=1,
|
62 |
+
error_correction=errCorr,
|
63 |
+
box_size=boxsize,
|
64 |
+
border=0,
|
65 |
+
)
|
66 |
+
qr.add_data(content)
|
67 |
+
qr.make(fit=True)
|
68 |
+
img = qr.make_image(fill_color="black", back_color="white")
|
69 |
+
|
70 |
+
# find smallest image size multiple of 256 that can fit qr
|
71 |
+
offset_min = 8 * boxsize
|
72 |
+
w, h = img.size
|
73 |
+
w = (w + 255 + offset_min) // 256 * 256
|
74 |
+
h = (h + 255 + offset_min) // 256 * 256
|
75 |
+
if w > 1024:
|
76 |
+
raise gr.Error("QR code is too large, please use a shorter content")
|
77 |
+
bg = Image.new('L', (w, h), 128)
|
78 |
+
|
79 |
+
# align on 16px grid
|
80 |
+
coords = ((w - img.size[0]) // 2 // boxsize * boxsize,
|
81 |
+
(h - img.size[1]) // 2 // boxsize * boxsize)
|
82 |
+
bg.paste(img, coords)
|
83 |
+
return bg
|
84 |
+
|
85 |
+
|
86 |
+
def inference(
|
87 |
+
qr_code_content: str,
|
88 |
+
errorCorrection: str,
|
89 |
+
prompt: str,
|
90 |
+
negative_prompt: str,
|
91 |
+
inferenceSteps: float,
|
92 |
+
guidance_scale: float = 10.0,
|
93 |
+
controlnet_conditioning_scale: float = 2.0,
|
94 |
+
seed: int = -1,
|
95 |
+
sampler="Euler a",
|
96 |
+
):
|
97 |
+
if prompt is None or prompt == "":
|
98 |
+
raise gr.Error("Prompt is required")
|
99 |
+
|
100 |
+
if qr_code_content is None or qr_code_content == "":
|
101 |
+
raise gr.Error("QR Code Content is required")
|
102 |
+
|
103 |
+
pipe.scheduler = SAMPLER_MAP[sampler](pipe.scheduler.config)
|
104 |
+
|
105 |
+
generator = torch.manual_seed(seed) if seed != -1 else torch.Generator()
|
106 |
+
|
107 |
+
print("Generating QR Code from content")
|
108 |
+
qrcode_image = create_code(qr_code_content, errorCorrection)
|
109 |
+
|
110 |
+
# hack due to gradio examples
|
111 |
+
init_image = qrcode_image
|
112 |
+
init_image.save("c:\\temp\\qr.jpg")
|
113 |
+
|
114 |
+
out = pipe(
|
115 |
+
prompt=prompt,
|
116 |
+
negative_prompt=negative_prompt,
|
117 |
+
image=qrcode_image,
|
118 |
+
width=qrcode_image.width,
|
119 |
+
height=qrcode_image.height,
|
120 |
+
guidance_scale=float(guidance_scale),
|
121 |
+
controlnet_conditioning_scale=float(controlnet_conditioning_scale),
|
122 |
+
generator=generator,
|
123 |
+
num_inference_steps=inferenceSteps,
|
124 |
+
)
|
125 |
+
return out.images[0]
|
126 |
+
|
127 |
+
|
128 |
+
css = """
|
129 |
+
#result_image {
|
130 |
+
display: flex;
|
131 |
+
place-content: center;
|
132 |
+
align-items: center;
|
133 |
+
}
|
134 |
+
#result_image > img {
|
135 |
+
height: auto;
|
136 |
+
max-width: 100%;
|
137 |
+
width: revert;
|
138 |
+
}
|
139 |
+
"""
|
140 |
+
|
141 |
+
with gr.Blocks(css=css) as blocks:
|
142 |
+
|
143 |
+
with gr.Row():
|
144 |
+
with gr.Column():
|
145 |
+
qr_code_content = gr.Textbox(
|
146 |
+
label="QR Code Content or URL",
|
147 |
+
info="The text you want to encode into the QR code",
|
148 |
+
value="",
|
149 |
+
)
|
150 |
+
errorCorrection = gr.Dropdown(
|
151 |
+
label="QR Code Error Correction Level",
|
152 |
+
choices=["L 7%", "M 15%", "Q 25%", "H 30%"],
|
153 |
+
value="H 30%"
|
154 |
+
)
|
155 |
+
|
156 |
+
prompt = gr.Textbox(
|
157 |
+
label="Prompt",
|
158 |
+
info="Prompt that guides the generation towards",
|
159 |
+
)
|
160 |
+
negative_prompt = gr.Textbox(
|
161 |
+
label="Negative Prompt",
|
162 |
+
value="ugly, disfigured, low quality, blurry, nsfw",
|
163 |
+
info="Prompt that guides the generation away from",
|
164 |
+
)
|
165 |
+
inferenceSteps = gr.Slider(
|
166 |
+
minimum=10.0,
|
167 |
+
maximum=60.0,
|
168 |
+
step=1,
|
169 |
+
value=20,
|
170 |
+
label="Inference Steps",
|
171 |
+
info="More steps give better image but longer runtime",
|
172 |
+
)
|
173 |
+
|
174 |
+
with gr.Accordion(
|
175 |
+
label="Params: The generated QR Code functionality is largely influenced by the parameters detailed below",
|
176 |
+
open=True,
|
177 |
+
):
|
178 |
+
controlnet_conditioning_scale = gr.Slider(
|
179 |
+
minimum=0.5,
|
180 |
+
maximum=2.5,
|
181 |
+
step=0.01,
|
182 |
+
value=1.5,
|
183 |
+
label="Controlnet Conditioning Scale",
|
184 |
+
info="""Controls the readability/creativity of the QR code.
|
185 |
+
High values: The generated QR code will be more readable.
|
186 |
+
Low values: The generated QR code will be more creative.
|
187 |
+
"""
|
188 |
+
)
|
189 |
+
guidance_scale = gr.Slider(
|
190 |
+
minimum=0.0,
|
191 |
+
maximum=25.0,
|
192 |
+
step=0.25,
|
193 |
+
value=7,
|
194 |
+
label="Guidance Scale",
|
195 |
+
info="Controls the amount of guidance the text prompt guides the image generation"
|
196 |
+
)
|
197 |
+
sampler = gr.Dropdown(choices=list(
|
198 |
+
SAMPLER_MAP.keys()), value="Euler a", label="Sampler")
|
199 |
+
seed = gr.Number(
|
200 |
+
minimum=-1,
|
201 |
+
maximum=9999999999,
|
202 |
+
value=-1,
|
203 |
+
label="Seed",
|
204 |
+
info="Seed for the random number generator. Set to -1 for a random seed"
|
205 |
+
)
|
206 |
+
with gr.Row():
|
207 |
+
run_btn = gr.Button("Run")
|
208 |
+
with gr.Column():
|
209 |
+
result_image = gr.Image(label="Result Image", elem_id="result_image")
|
210 |
+
run_btn.click(
|
211 |
+
inference,
|
212 |
+
inputs=[
|
213 |
+
qr_code_content,
|
214 |
+
errorCorrection,
|
215 |
+
prompt,
|
216 |
+
negative_prompt,
|
217 |
+
inferenceSteps,
|
218 |
+
guidance_scale,
|
219 |
+
controlnet_conditioning_scale,
|
220 |
+
seed,
|
221 |
+
sampler,
|
222 |
+
],
|
223 |
+
outputs=[result_image],
|
224 |
+
)
|
225 |
+
|
226 |
+
blocks.queue(concurrency_count=1, max_size=20, api_open=False)
|
227 |
+
blocks.launch(share=bool(os.environ.get("SHARE", True)), show_api=False)
|