Spaces:
Sleeping
Sleeping
File size: 9,531 Bytes
ad4deb5 a29bf9d 9621967 d7137df a29bf9d ad4deb5 9621967 ad4deb5 | 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 | import gradio as gr
import numpy as np
from numba import njit
from PIL import Image, ImageOps
import io
import tempfile
import os
from huggingface_hub import hf_hub_download
files = ["a.jpg", "b.png", "c.png", 'd.png']
paths = [
hf_hub_download(
repo_id="Rothfeld/drostescher",
filename=f,
repo_type="space"
)
for f in files
]
import shutil
for a,b in zip(files,paths):
shutil.copy(b,a)
A,B,C,D = files
# ββ forward transformation ββββββββββββββββββββββββββββββββββββββββββββββββββ
@njit(fastmath=True)
def tl_to_yx(t, l, w, h):
y = (1 - t / h) * 2 - 1
x = (l / w) * 2 - 1
return y, x
@njit(fastmath=True)
def yx_to_ra(y, x):
a = np.arctan2(y, x) % (2 * np.pi)
r = (x * x + y * y) ** 0.5
return r, a
@njit(fastmath=True)
def a_to_na(a):
return a / (2 * np.pi)
@njit(fastmath=True)
def r_to_logr(r):
return np.log2(r)
@njit(fastmath=True)
def encode_to_logrna(l, t, w, h):
y, x = tl_to_yx(t, l, w, h)
r, a = yx_to_ra(y, x)
logr = r_to_logr(r)
na = a_to_na(a)
return logr, na
# ββ inverse transformation ββββββββββββββββββββββββββββββββββββββββββββββββββ
@njit(fastmath=True)
def na_to_a(na):
return na * (2 * np.pi)
@njit(fastmath=True)
def logr_to_r(logr):
return 2 ** logr
@njit(fastmath=True)
def ra_to_yx(r, a):
x = r * np.cos(a)
y = r * np.sin(a)
return y, x
@njit(fastmath=True)
def yx_to_tl(y, x, w, h):
t = (1 - (y + 1) / 2) * h
l = ((x + 1) / 2) * w
t = np.round(t).astype(np.int64)
l = np.round(l).astype(np.int64)
return t, l
@njit(fastmath=True)
def decode_to_tl(logr, na, w, h):
a = na_to_a(na)
r = logr_to_r(logr)
y, x = ra_to_yx(r, a)
t, l = yx_to_tl(y, x, w, h)
t %= h
l %= w
return t, l
# ββ utilities βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def drostify(o: Image.Image):
w, h = o.width, o.height
scale = 1 / 2
w2, h2 = int(w * scale), int(h * scale)
hb = (w - w2) // 2
vb = (h - h2) // 2
l, t, r, b = (hb, vb, w - hb, h - vb)
small = o.resize((r - l, b - t))
o.paste(small, (l, t, r, b), mask=small.getchannel('A'))
def pad_to_aspect(img, target_ratio, fill=0):
w, h = img.size
current_ratio = w / h
if current_ratio > target_ratio:
new_h = int(round(w / target_ratio))
pad_total = new_h - h
top = pad_total // 2
bottom = pad_total - top
left = right = 0
else:
new_w = int(round(h * target_ratio))
pad_total = new_w - w
left = pad_total // 2
right = pad_total - left
top = bottom = 0
return ImageOps.expand(img, (left, top, right, bottom), fill=fill)
def make_still(img_pil, output_size):
"""Single Droste-effect still frame."""
o = img_pil.convert("RGBA")
o = pad_to_aspect(o, output_size[0] / output_size[1])
o = o.resize(output_size, Image.Resampling.NEAREST)
drostify(o)
w, h = o.width, o.height
t, l = np.meshgrid(np.arange(h), np.arange(w), indexing="ij")
logr, na = encode_to_logrna(l, t, w, h)
logr -= na
logr %= -1
t2, l2 = decode_to_tl(logr, na, w, h)
oa = np.array(o)
ia = oa[t2, l2]
for i in range(1, 10):
transparent = ia[:, :, 3] == 0
if not transparent.any():
break
dt, dl = decode_to_tl(logr[transparent] - i, na[transparent], w, h)
ia[transparent] = oa[dt, dl]
return Image.fromarray(ia)
def make_animation(img_pil, output_size, n_frames, n_rotations):
"""Animated Droste zoom loop."""
o = img_pil.convert("RGBA")
o = pad_to_aspect(o, output_size[0] / output_size[1])
o = o.resize(output_size, Image.Resampling.NEAREST)
drostify(o)
origa = np.array(o.copy())
w, h = o.width, o.height
t0, l0 = np.meshgrid(np.arange(h), np.arange(w), indexing="ij")
steps = np.linspace(0.0, 1.0, n_frames, endpoint=False)
frames = []
for s in steps:
logr, na = encode_to_logrna(l0, t0, w, h)
logr -= s
logr -= na
logr %= -1
nac = na + s * n_rotations
t2, l2 = decode_to_tl(logr, nac, w, h)
ia = origa[t2, l2]
for i in range(1, 5):
transparent = ia[:, :, 3] == 0
if not transparent.any():
break
dt, dl = decode_to_tl(logr[transparent] - i, nac[transparent], w, h)
ia[transparent] = origa[dt, dl]
frames.append(Image.fromarray(ia).convert("RGB"))
return frames
# ββ Gradio callbacks ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
def run_still(image, width, height):
if image is None:
return None, "β οΈ Please upload an image."
try:
pil = Image.fromarray(image)
result = make_still(pil, (int(width), int(height)))
return np.array(result.convert("RGB")), "β
Done!"
except Exception as e:
return None, f"β Error: {e}"
def run_animation(image, width, height, n_frames, n_rotations, fps):
if image is None:
return None, "β οΈ Please upload an image."
try:
pil = Image.fromarray(image)
frames = make_animation(pil, (int(width), int(height)), int(n_frames), int(n_rotations))
tmp = tempfile.NamedTemporaryFile(suffix=".gif", delete=False)
frames[0].save(
tmp.name,
save_all=True,
append_images=frames[1:],
duration=int(1000 / fps),
loop=0,
disposal=2,
)
return tmp.name, f"β
{len(frames)} frames @ {fps} fps"
except Exception as e:
return None, f"β Error: {e}"
# ββ UI βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
DESCRIPTION = """
# π Droste Effect
Upload any image (transparency supported) and apply the **Droste effect** β
an infinite self-similar zoom that maps the image into a logarithmic spiral.
Two modes:
- **Still** β a single warped frame
- **Animation** β a seamlessly looping GIF zoom
"""
with gr.Blocks(title="Droste Effect") as demo:
gr.Markdown(DESCRIPTION)
with gr.Row():
with gr.Column(scale=1):
image_input = gr.Image(label="Input image", type="numpy", image_mode='RGBA')
with gr.Accordion("Output size", open=False):
width = gr.Slider(64, 1024, value=400, step=8, label="Width")
height = gr.Slider(64, 1024, value=400, step=8, label="Height")
with gr.Column(scale=2):
with gr.Tab("πΌοΈ Still"):
still_btn = gr.Button("Generate still", variant="primary")
still_out = gr.Image(label="Droste still", type="numpy", format='png')
still_status = gr.Textbox(label="Status", interactive=False, max_lines=1)
still_btn.click(
run_still,
inputs=[image_input, width, height],
outputs=[still_out, still_status],
)
with gr.Tab("ποΈ Animation"):
with gr.Row():
n_frames = gr.Slider(12, 120, value=60, step=4, label="Frames")
n_rotations = gr.Slider(0, 8, value=0, step=1, label="Rotations per loop")
fps = gr.Slider(6, 60, value=30, step=2, label="FPS")
anim_btn = gr.Button("Generate animation", variant="primary")
anim_out = gr.Image(label="Droste animation (GIF)")
anim_status = gr.Textbox(label="Status", interactive=False, max_lines=1)
anim_btn.click(
run_animation,
inputs=[image_input, width, height, n_frames, n_rotations, fps],
outputs=[anim_out, anim_status],
)
# ββ Examples βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
gr.Markdown("### πΌοΈ Example images β click to load")
# Still examples: [image, width, height]
gr.Examples(
examples=[
[A, 1024, 1024],
[B, 400, 400],
[C, 512, 512],
[D, 400, 400],
],
inputs=[image_input, width, height],
outputs=[still_out, still_status],
fn=run_still,
cache_examples=False,
label="Still examples",
)
gr.Markdown("""
---
**How it works:** each pixel's Cartesian coordinates are converted to polar form,
the radius is logβ-transformed, and the angle is used to modulate the radius offset,
turning circles into spirals. The image is then resampled in that warped space,
creating the recursive Droste zoom. Transparent pixels are resolved by stepping
one ring inward until opaque colour is found.
Space generated from ./_orig.py using claude.
""")
if __name__ == "__main__":
demo.launch() |