Spaces:
Sleeping
Sleeping
File size: 5,675 Bytes
ec4da21 | 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 | from __future__ import annotations
from pathlib import Path
from typing import Dict, Tuple
import numpy as np
from PIL import Image, ImageFilter, ImageStat, UnidentifiedImageError
from config import (
ALMOST_BLACK_MEAN,
ALMOST_WHITE_MEAN,
LOW_STDDEV,
MIN_IMAGE_SIZE,
)
def _laplacian_like_sharpness(gray: Image.Image) -> float:
edges = gray.filter(ImageFilter.FIND_EDGES)
arr = np.asarray(edges, dtype=np.float32)
return float(arr.var())
def _edge_density(gray: Image.Image) -> float:
edges = gray.filter(ImageFilter.FIND_EDGES)
arr = np.asarray(edges, dtype=np.float32)
threshold = arr.mean() + arr.std()
if threshold <= 0:
return 0.0
density = (arr > threshold).mean()
return float(density)
def _dominant_colors_count(img: Image.Image, max_colors: int = 12) -> int:
small = img.convert("RGB").resize((64, 64))
palette_img = small.quantize(colors=max_colors, method=Image.MEDIANCUT)
colors = palette_img.getcolors()
return len(colors) if colors else 0
def _whitespace_ratio(gray: Image.Image) -> float:
arr = np.asarray(gray, dtype=np.float32)
near_white = (arr > 240).mean()
near_black = (arr < 15).mean()
return float(max(near_white, near_black))
def _layout_density(gray: Image.Image) -> float:
edges = gray.filter(ImageFilter.FIND_EDGES)
arr = np.asarray(edges, dtype=np.float32)
active = (arr > 30).mean()
return float(active)
def _center_activity(gray: Image.Image) -> float:
arr = np.asarray(gray.filter(ImageFilter.FIND_EDGES), dtype=np.float32)
h, w = arr.shape
y1, y2 = int(h * 0.25), int(h * 0.75)
x1, x2 = int(w * 0.25), int(w * 0.75)
center = arr[y1:y2, x1:x2]
if center.size == 0:
return 0.0
return float((center > 30).mean())
def _grid_balance_3x3(gray: Image.Image) -> float:
arr = np.asarray(gray.filter(ImageFilter.FIND_EDGES), dtype=np.float32)
h, w = arr.shape
ys = np.linspace(0, h, 4, dtype=int)
xs = np.linspace(0, w, 4, dtype=int)
cells = []
for i in range(3):
for j in range(3):
cell = arr[ys[i]:ys[i + 1], xs[j]:xs[j + 1]]
if cell.size == 0:
cells.append(0.0)
else:
cells.append(float((cell > 30).mean()))
mean_val = float(np.mean(cells))
std_val = float(np.std(cells))
balance = max(0.0, 1.0 - std_val / (mean_val + 1e-6))
return balance
def inspect_image_content(image_path: Path) -> Tuple[bool, str]:
try:
with Image.open(image_path) as img:
img.load()
width, height = img.size
if width < MIN_IMAGE_SIZE or height < MIN_IMAGE_SIZE:
return False, f"too_small_{width}x{height}"
gray = img.convert("L")
extrema = gray.getextrema()
if extrema is None:
return False, "failed_extrema_check"
if extrema[0] == extrema[1]:
return False, "blank_uniform_image"
stat = ImageStat.Stat(gray)
mean_val = stat.mean[0]
stddev = stat.stddev[0]
if mean_val > ALMOST_WHITE_MEAN and stddev < LOW_STDDEV:
return False, "almost_blank_white_image"
if mean_val < ALMOST_BLACK_MEAN and stddev < LOW_STDDEV:
return False, "almost_blank_black_image"
return True, "ok"
except UnidentifiedImageError:
return False, "unidentified_image"
except Exception as e:
return False, f"image_inspection_error: {e}"
def extract_features(image_path: Path) -> Dict[str, float | int | bool | str]:
has_content, reason = inspect_image_content(image_path)
if not has_content:
return {
"content_present_rule": False,
"blank_reason": reason,
"mean_brightness": 0.0,
"contrast": 0.0,
"saturation_mean": 0.0,
"dominant_colors_count": 0,
"sharpness": 0.0,
"edge_density": 0.0,
"whitespace_ratio": 1.0,
"layout_density": 0.0,
"center_activity": 0.0,
"grid_balance_3x3": 0.0,
}
with Image.open(image_path) as img:
img = img.convert("RGB")
gray = img.convert("L")
hsv = img.convert("HSV")
gray_stat = ImageStat.Stat(gray)
hsv_stat = ImageStat.Stat(hsv)
mean_brightness = float(gray_stat.mean[0]) / 255.0
contrast = float(gray_stat.stddev[0]) / 64.0
saturation_mean = float(hsv_stat.mean[1]) / 255.0
dominant_colors_count = _dominant_colors_count(img)
sharpness = _laplacian_like_sharpness(gray) / 1000.0
edge_density = _edge_density(gray)
whitespace_ratio = _whitespace_ratio(gray)
layout_density = _layout_density(gray)
center_activity = _center_activity(gray)
grid_balance_3x3 = _grid_balance_3x3(gray)
return {
"content_present_rule": True,
"blank_reason": "ok",
"mean_brightness": mean_brightness,
"contrast": contrast,
"saturation_mean": saturation_mean,
"dominant_colors_count": dominant_colors_count,
"sharpness": sharpness,
"edge_density": edge_density,
"whitespace_ratio": whitespace_ratio,
"layout_density": layout_density,
"center_activity": center_activity,
"grid_balance_3x3": grid_balance_3x3,
} |