File size: 13,940 Bytes
823039e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 | """
Vector Studio — PNG/JPG to SVG
Clean line work for generated images.
Hugging Face Space (Gradio SDK). Run locally: python app.py
"""
import os
import re
import base64
import tempfile
import uuid
import numpy as np
from PIL import Image, ImageOps, ImageFilter
import gradio as gr
import vtracer
MAX_SIDE = 2600 # Cap after upscaling, keeps free CPU Spaces from timing out
OUT_DIR = os.path.join(tempfile.gettempdir(), "svg_out")
os.makedirs(OUT_DIR, exist_ok=True)
# --------------------------------------------------------------------------
# Image preparation
# --------------------------------------------------------------------------
def flatten(img: Image.Image) -> Image.Image:
"""Composite alpha onto white so transparency isn't traced as black."""
img = ImageOps.exif_transpose(img)
if img.mode in ("RGBA", "LA", "P"):
img = img.convert("RGBA")
bg = Image.new("RGBA", img.size, (255, 255, 255, 255))
img = Image.alpha_composite(bg, img)
return img.convert("RGB")
def otsu_threshold(arr: np.ndarray) -> int:
hist = np.bincount(arr.ravel(), minlength=256).astype(np.float64)
total = hist.sum()
omega = np.cumsum(hist) / total
mu = np.cumsum(hist * np.arange(256)) / total
mu_t = mu[-1]
denom = omega * (1.0 - omega)
denom[denom == 0] = 1e-12
sigma_b = (mu_t * omega - mu) ** 2 / denom
return int(np.argmax(sigma_b))
def binarize(gray: Image.Image, method: str, level: int, block: int, offset: int) -> np.ndarray:
"""Return a boolean mask where True means ink."""
arr = np.asarray(gray, dtype=np.uint8)
if method.startswith("Adaptive"):
radius = max(1, int(block) // 2)
local = np.asarray(gray.filter(ImageFilter.BoxBlur(radius)), dtype=np.int16)
return arr.astype(np.int16) < (local - int(offset))
if method.startswith("Otsu"):
level = otsu_threshold(arr)
return arr < int(level)
def zhang_suen(mask: np.ndarray, max_iter: int = 60) -> np.ndarray:
"""Skeletonize: thin every stroke down to a 1 px centerline."""
img = mask.copy()
for _ in range(max_iter):
changed = False
for step in (0, 1):
P = np.pad(img, 1, constant_values=False)
P2, P3, P4 = P[:-2, 1:-1], P[:-2, 2:], P[1:-1, 2:]
P5, P6, P7 = P[2:, 2:], P[2:, 1:-1], P[2:, :-2]
P8, P9 = P[1:-1, :-2], P[:-2, :-2]
ring = [P2, P3, P4, P5, P6, P7, P8, P9, P2]
B = sum(x.astype(np.uint8) for x in ring[:8])
A = np.zeros(img.shape, np.uint8)
for i in range(8):
A += (~ring[i] & ring[i + 1]).astype(np.uint8)
base = img & (B >= 2) & (B <= 6) & (A == 1)
if step == 0:
cond = base & ~(P2 & P4 & P6) & ~(P4 & P6 & P8)
else:
cond = base & ~(P2 & P4 & P8) & ~(P2 & P6 & P8)
if cond.any():
img &= ~cond
changed = True
if not changed:
break
return img
def set_stroke_width(mask: np.ndarray, width: int) -> np.ndarray:
"""Grow the skeleton back out to one uniform stroke width."""
if width <= 1:
return mask
ink = Image.fromarray(np.where(mask, 255, 0).astype(np.uint8))
size = width if width % 2 == 1 else width + 1
ink = ink.filter(ImageFilter.MaxFilter(size))
return np.asarray(ink) > 127
def cap_size(img: Image.Image) -> Image.Image:
w, h = img.size
if max(w, h) <= MAX_SIDE:
return img
s = MAX_SIDE / max(w, h)
return img.resize((max(1, int(w * s)), max(1, int(h * s))), Image.LANCZOS)
# --------------------------------------------------------------------------
# SVG post-processing
# --------------------------------------------------------------------------
def strip_background(svg: str) -> str:
"""Drop near-white fills, which is what the background becomes in color mode."""
def is_light(hexcol: str) -> bool:
r, g, b = (int(hexcol[i:i + 2], 16) for i in (1, 3, 5))
return r > 243 and g > 243 and b > 243
return re.sub(
r'<path[^>]*fill="(#[0-9a-fA-F]{6})"[^>]*/>\s*',
lambda m: "" if is_light(m.group(1)) else m.group(0),
svg,
)
def recolor(svg: str, color: str) -> str:
return re.sub(r'fill="#000000"', f'fill="{color}"', svg)
def rescale_root(svg: str, out_w: int, out_h: int) -> str:
"""Add a viewBox and set the display size back to the source dimensions."""
m = re.search(r'<svg([^>]*)width="(\d+)"\s+height="(\d+)"', svg)
if not m:
return svg
vw, vh = m.group(2), m.group(3)
new_tag = (
f'<svg{m.group(1)}width="{out_w}" height="{out_h}" '
f'viewBox="0 0 {vw} {vh}"'
)
return svg[: m.start()] + new_tag + svg[m.end():]
def preview_html(svg: str, label: str) -> str:
if len(svg) > 4_000_000:
return (
f"<div class='vs-note'>{label} is large ({len(svg) // 1024} KB). "
"Download it instead of previewing here, or raise "
"<em>Filter small specks</em> to cut the path count.</div>"
)
b64 = base64.b64encode(svg.encode("utf-8")).decode("ascii")
return (
"<div class='vs-stage'>"
f"<img src='data:image/svg+xml;base64,{b64}' alt='{label}'/>"
"</div>"
)
# --------------------------------------------------------------------------
# Main pipeline
# --------------------------------------------------------------------------
def vectorize(
image,
mode,
invert,
thr_method,
thr_level,
block,
offset,
denoise,
upscale,
uniform,
stroke_w,
curve_mode,
speckle,
corner,
length_thr,
splice,
precision,
colors,
layer_diff,
hierarchical,
drop_bg,
line_color,
progress=gr.Progress(),
):
if image is None:
raise gr.Error("Upload a PNG or JPG first.")
progress(0.1, desc="Preparing image")
src = flatten(image)
orig_w, orig_h = src.size
work = src
if upscale > 1:
work = work.resize(
(int(orig_w * upscale), int(orig_h * upscale)), Image.LANCZOS
)
work = cap_size(work)
line_mode = mode.startswith("Line")
if line_mode:
progress(0.3, desc="Isolating lines")
gray = ImageOps.autocontrast(work.convert("L"), cutoff=1)
if invert:
gray = ImageOps.invert(gray)
if denoise > 0:
gray = gray.filter(ImageFilter.MedianFilter(int(denoise) * 2 + 1))
mask = binarize(gray, thr_method, thr_level, block, offset)
if uniform:
progress(0.45, desc="Finding centerlines")
mask = set_stroke_width(zhang_suen(mask), int(stroke_w))
prepped = Image.fromarray(np.where(mask, 0, 255).astype(np.uint8)).convert("RGB")
else:
progress(0.35, desc="Reducing color areas")
prepped = work
if denoise > 0:
prepped = prepped.filter(ImageFilter.MedianFilter(int(denoise) * 2 + 1))
stem = uuid.uuid4().hex[:10]
tmp_png = os.path.join(OUT_DIR, f"{stem}.png")
out_svg = os.path.join(OUT_DIR, "vector.svg" if line_mode else "vector-color.svg")
prepped.save(tmp_png)
progress(0.6, desc="Tracing paths")
vtracer.convert_image_to_svg_py(
tmp_png,
out_svg,
colormode="binary" if line_mode else "color",
hierarchical="cutout" if hierarchical == "Cutout" else "stacked",
mode="polygon" if curve_mode.startswith("Polygon") else "spline",
filter_speckle=int(speckle),
color_precision=int(colors),
layer_difference=int(layer_diff),
corner_threshold=int(corner),
length_threshold=float(length_thr),
max_iterations=10,
splice_threshold=int(splice),
path_precision=int(precision),
)
progress(0.85, desc="Cleaning up SVG")
svg = open(out_svg, "r", encoding="utf-8").read()
if drop_bg:
svg = strip_background(svg)
if line_mode:
svg = recolor(svg, line_color)
svg = rescale_root(svg, orig_w, orig_h)
with open(out_svg, "w", encoding="utf-8") as f:
f.write(svg)
try:
os.remove(tmp_png)
except OSError:
pass
n_paths = svg.count("<path")
stats = (
f"**{n_paths} path{'' if n_paths == 1 else 's'}** · {len(svg) / 1024:.1f} KB · "
f"canvas {orig_w}×{orig_h} px · traced at {prepped.size[0]}×{prepped.size[1]} px"
)
return preview_html(svg, "Vector result"), out_svg, stats
# --------------------------------------------------------------------------
# Interface
# --------------------------------------------------------------------------
CSS = """
.vs-stage {
background-color: #f7f5f2;
background-image:
linear-gradient(45deg, #e6e2dc 25%, transparent 25%, transparent 75%, #e6e2dc 75%),
linear-gradient(45deg, #e6e2dc 25%, transparent 25%, transparent 75%, #e6e2dc 75%);
background-size: 18px 18px;
background-position: 0 0, 9px 9px;
border-radius: 10px; padding: 14px; min-height: 240px;
display: flex; align-items: center; justify-content: center;
}
.vs-stage img { max-width: 100%; max-height: 62vh; }
.vs-note { padding: 18px; font-size: 0.9rem; line-height: 1.5; }
footer { display: none !important; }
"""
# Gradio 6 wants theme/css on launch(), Gradio 5 on the Blocks constructor.
_MAJOR = int(gr.__version__.split(".")[0])
_STYLE = {"theme": gr.themes.Soft(), "css": CSS}
_BLOCKS_KW = {} if _MAJOR >= 6 else _STYLE
_LAUNCH_KW = _STYLE if _MAJOR >= 6 else {}
with gr.Blocks(title="PNG/JPG to SVG", **_BLOCKS_KW) as demo:
gr.Markdown(
"## PNG / JPG to SVG\n"
"Turn bitmaps into clean, scalable paths. "
"Built for AI-generated drawings, logos and sketches."
)
with gr.Row():
# ---------------- left column: input and controls ----------------
with gr.Column(scale=4):
image = gr.Image(label="Image", type="pil", sources=["upload", "clipboard"],
image_mode="RGBA", height=300)
mode = gr.Radio(
["Line art (black & white)", "Color (filled shapes)"],
value="Line art (black & white)",
label="Mode",
)
run = gr.Button("Vectorize", variant="primary")
with gr.Accordion("Isolate lines", open=True) as line_box:
thr_method = gr.Radio(
["Otsu (automatic)", "Global", "Adaptive (uneven lighting)"],
value="Otsu (automatic)",
label="Threshold",
)
thr_level = gr.Slider(0, 255, 128, step=1, label="Cutoff (Global only)")
block = gr.Slider(5, 151, 41, step=2, label="Window size (Adaptive only)")
offset = gr.Slider(0, 40, 8, step=1, label="Sensitivity (Adaptive only)")
invert = gr.Checkbox(False, label="Invert (light lines on dark background)")
denoise = gr.Slider(0, 3, 1, step=1, label="Smooth noise")
upscale = gr.Slider(1, 4, 2, step=1,
label="Upscale before tracing (smoother curves)")
uniform = gr.Checkbox(
False, label="Uniform stroke width (redraw from centerlines)"
)
stroke_w = gr.Slider(1, 9, 3, step=2, label="Stroke width in px")
with gr.Accordion("Path quality", open=False):
curve_mode = gr.Radio(
["Spline (smooth curves)", "Polygon (hard edges)"],
value="Spline (smooth curves)", label="Curve type",
)
speckle = gr.Slider(0, 40, 6, step=1, label="Filter small specks")
corner = gr.Slider(0, 180, 60, step=1, label="Corner threshold")
length_thr = gr.Slider(0.5, 10, 4, step=0.5, label="Shortest segment")
splice = gr.Slider(0, 180, 45, step=1, label="Curve splicing")
precision = gr.Slider(1, 8, 4, step=1, label="Coordinate precision")
with gr.Accordion("Color and output", open=False):
colors = gr.Slider(1, 8, 6, step=1, label="Color depth (color mode)")
layer_diff = gr.Slider(0, 64, 16, step=1, label="Layer spacing (color mode)")
hierarchical = gr.Radio(
["Stacked", "Cutout"], value="Stacked", label="Layer structure"
)
drop_bg = gr.Checkbox(True, label="Remove white background")
line_color = gr.ColorPicker("#111111", label="Line color")
# ---------------- right column: result ----------------
with gr.Column(scale=6):
preview = gr.HTML("<div class='vs-stage'><span class='vs-note'>"
"Your result appears here.</span></div>",
label="Preview")
stats = gr.Markdown("")
download = gr.File(label="Download SVG", height=90)
gr.Markdown(
"**For clean lines:** upscale 2–3×, use spline curves, set speck filtering to 6–12. "
"If a drawing comes out ragged, turn on *Uniform stroke width* — it pulls every "
"line onto a centerline and redraws it at a constant weight."
)
def toggle(m):
return gr.update(open=m.startswith("Line"))
mode.change(toggle, mode, line_box)
inputs = [image, mode, invert, thr_method, thr_level, block, offset, denoise,
upscale, uniform, stroke_w, curve_mode, speckle, corner, length_thr,
splice, precision, colors, layer_diff, hierarchical, drop_bg, line_color]
run.click(vectorize, inputs, [preview, download, stats])
if __name__ == "__main__":
demo.queue(max_size=20).launch(
server_name="0.0.0.0",
server_port=int(os.environ.get("PORT", 7860)),
ssr_mode=False,
**_LAUNCH_KW,
)
|