Spaces:
Sleeping
Sleeping
File size: 12,912 Bytes
e7f1d57 |
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 |
# Dependencies
import numpy as np
from scipy.stats import entropy
from utils.logger import get_logger
from config.schemas import MetricResult
from config.constants import MetricType
from utils.image_processor import ImageProcessor
from config.constants import TEXTURE_ANALYSIS_PARAMS
# Suppress NumPy warning
np.seterr(divide = 'ignore',
invalid = 'ignore',
)
# Setup Logging
logger = get_logger(__name__)
class TextureAnalyzer:
"""
Statistical texture analysis for AI detection
Core principle:
---------------
- Real photos : Natural texture variation (random but structured)
- AI images : Either too smooth or repetitive patterns
Method:
-------
1. Extract local patches
2. Compute texture features (contrast, entropy)
3. Analyze texture consistency and distribution
4. Detect unnaturally smooth regions
"""
def __init__(self):
"""
Initialize TextureAnalyzer Class
"""
self.patch_size = TEXTURE_ANALYSIS_PARAMS.PATCH_SIZE
self.n_patches = TEXTURE_ANALYSIS_PARAMS.N_PATCHES
self.image_processor = ImageProcessor()
self._rng = np.random.default_rng(seed = TEXTURE_ANALYSIS_PARAMS.RANDOM_SEED)
def detect(self, image: np.ndarray) -> MetricResult:
"""
Run texture analysis
Arguments:
----------
image { np.ndarray } : RGB image array (H, W, 3)
Returns:
--------
{ MetricResult } : Structured Texture-domain metric result containing:
- score : Suspicion score [0.0, 1.0]
- confidence : Reliability of texture evidence
- details : Texture forensics and statistics
"""
try:
logger.debug(f"Running texture analysis on image shape {image.shape}")
# Convert to luminance
luminance = self.image_processor.rgb_to_luminance(image = image)
# Extract patches
patches = self._extract_patches(luminance = luminance)
if (len(patches) == 0):
logger.warning("No patches extracted for texture analysis")
return MetricResult(metric_type = MetricType.TEXTURE,
score = TEXTURE_ANALYSIS_PARAMS.NEUTRAL_SCORE,
confidence = 0.0,
details = {"reason": "no_patches_extracted"},
)
# Compute texture features
texture_features, texture_metadata = self._compute_texture_features(patches = patches)
# Analyze for anomalies
texture_score, texture_details = self._analyze_texture_anomalies(features = texture_features,
metadata = texture_metadata,
)
# Calculate Confidence
confidence = float(np.clip((abs(texture_score - TEXTURE_ANALYSIS_PARAMS.NEUTRAL_SCORE) * 2.0), 0.0, 1.0))
logger.debug(f"Texture analysis: Texture Score={texture_score:.3f}, patches={len(patches)}")
return MetricResult(metric_type = MetricType.TEXTURE,
score = float(texture_score),
confidence = confidence,
details = {"patches_total" : int(len(patches)),
**texture_metadata,
**texture_details,
},
)
except Exception as e:
logger.error(f"Texture analysis failed: {e}")
# Return neutral score on error
return MetricResult(metric_type = MetricType.TEXTURE,
score = TEXTURE_ANALYSIS_PARAMS.NEUTRAL_SCORE,
confidence = 0.0,
details = {"error": "texture_analysis_failed"},
)
def _extract_patches(self, luminance: np.ndarray) -> np.ndarray:
"""
Extract random patches from image
"""
h, w = luminance.shape
if ((h < self.patch_size) or (w < self.patch_size)):
logger.warning(f"Image too small for patch size {self.patch_size}")
return np.array([])
patches = list()
for _ in range(self.n_patches):
y = self._rng.integers(0, h - self.patch_size)
x = self._rng.integers(0, w - self.patch_size)
patch = luminance[y:y+self.patch_size, x:x+self.patch_size]
patches.append(patch)
return np.array(patches)
def _compute_texture_features(self, patches: np.ndarray) -> tuple[dict, dict]:
"""
Compute texture features for each patch
Features:
---------
1. Local contrast (standard deviation)
2. Entropy (randomness)
3. Smoothness (inverse of variance)
4. Edge density
Arguments:
----------
patches { np.ndarray } : Array of patches
Returns:
--------
{ tuple } : A tuple containing
- A dictionary of feature arrays
- A dictionary of texture analysis metadata
"""
contrasts = list()
entropies = list()
smoothnesses = list()
edge_densities = list()
uniform_skipped = 0
for patch in patches:
pmin = patch.min()
pmax = patch.max()
if ((pmax - pmin < 1e-6)):
# skip fully uniform patch entirely
uniform_skipped += 1
continue
# Contrast (std deviation)
contrast = np.std(patch)
contrasts.append(contrast)
# Entropy (using histogram)
hist, _ = np.histogram(patch,
bins = TEXTURE_ANALYSIS_PARAMS.HISTOGRAM_BINS,
range = TEXTURE_ANALYSIS_PARAMS.HISTOGRAM_RANGE,
)
hist = hist / (np.sum(hist) + 1e-10)
ent = entropy(hist + 1e-10)
entropies.append(ent)
# Smoothness (inverse of variance, scaled)
variance = np.var(patch)
smoothness = 1.0 / (1.0 + variance)
smoothnesses.append(smoothness)
# Edge density (using Sobel)
gx, gy = self.image_processor.compute_gradients(luminance = patch)
gradient_mag = np.sqrt(gx**2 + gy**2)
edge_density = np.mean(gradient_mag > TEXTURE_ANALYSIS_PARAMS.EDGE_THRESHOLD)
edge_densities.append(edge_density)
# Construct results in proper format
features = {"contrast" : np.array(contrasts),
"entropy" : np.array(entropies),
"smoothness" : np.array(smoothnesses),
"edge_density" : np.array(edge_densities),
}
metadata = {"patches_used" : int(len(contrasts)),
"uniform_patches_skipped" : int(uniform_skipped),
}
return features, metadata
def _analyze_texture_anomalies(self, features: dict, metadata: dict) -> tuple[float, dict]:
"""
Analyze texture features for AI generation artifacts
Checks:
-------
1. Excessive smoothness (too many overly smooth patches)
2. Entropy distribution (too uniform = suspicious)
3. Contrast consistency
Arguments:
----------
features { dict } : Dictionary of texture features
metadata { dict } : Dictionary of texture analysis metadata
Returns:
--------
{ tuple } : A tuple containing:
- Suspicion score [0.0, 1.0]
- Texture statistics
"""
contrast = features['contrast']
entropy_vals = features['entropy']
smoothness = features['smoothness']
edge_density = features['edge_density']
if ((len(contrast) == 0) or (len(entropy_vals) == 0) or (len(smoothness) == 0) or (len(edge_density) == 0)):
logger.debug("All texture features filtered out; returning neutral score")
return (TEXTURE_ANALYSIS_PARAMS.NEUTRAL_SCORE,
{"reason": "all_texture_features_filtered"},
)
# Early exit: all patches nearly uniform
if (np.all(contrast < 1e-6)):
logger.debug("All texture patches near-uniform; returning neutral score")
return (TEXTURE_ANALYSIS_PARAMS.NEUTRAL_SCORE,
{"reason": "all_patches_near_uniform"},
)
# Smoothness Analysis
smooth_ratio = np.mean(smoothness > TEXTURE_ANALYSIS_PARAMS.SMOOTHNESS_THRESHOLD)
smoothness_anomaly = 0.0
if (smooth_ratio > TEXTURE_ANALYSIS_PARAMS.SMOOTH_RATIO_THRESHOLD):
# More than 40% very smooth patches
smoothness_anomaly = min(1.0, (smooth_ratio - TEXTURE_ANALYSIS_PARAMS.SMOOTH_RATIO_THRESHOLD) * TEXTURE_ANALYSIS_PARAMS.SMOOTH_RATIO_SCALE)
# Entropy distribution Analysis
entropy_cv = np.std(entropy_vals) / (np.mean(entropy_vals) + 1e-10)
entropy_anomaly = 0.0
if (entropy_cv < TEXTURE_ANALYSIS_PARAMS.ENTROPY_CV_THRESHOLD):
# Too uniform
entropy_anomaly = (TEXTURE_ANALYSIS_PARAMS.ENTROPY_CV_THRESHOLD - entropy_cv) * TEXTURE_ANALYSIS_PARAMS.ENTROPY_SCALE
# Contrast distribution Analysis
contrast_cv = np.std(contrast) / (np.mean(contrast) + 1e-10)
contrast_anomaly = 0.0
if (contrast_cv < TEXTURE_ANALYSIS_PARAMS.CONTRAST_CV_LOW):
# Too uniform
contrast_anomaly = (TEXTURE_ANALYSIS_PARAMS.CONTRAST_CV_LOW - contrast_cv) * TEXTURE_ANALYSIS_PARAMS.CONTRAST_LOW_SCALE
elif (contrast_cv > TEXTURE_ANALYSIS_PARAMS.CONTRAST_CV_HIGH):
# Too variable (suspicious)
contrast_anomaly = min(1.0, (contrast_cv - TEXTURE_ANALYSIS_PARAMS.CONTRAST_CV_HIGH) * TEXTURE_ANALYSIS_PARAMS.CONTRAST_HIGH_SCALE)
# Edge density consistency Analysis
edge_cv = np.std(edge_density) / (np.mean(edge_density) + 1e-10)
edge_anomaly = 0.0
if (edge_cv < TEXTURE_ANALYSIS_PARAMS.EDGE_CV_THRESHOLD):
edge_anomaly = (TEXTURE_ANALYSIS_PARAMS.EDGE_CV_THRESHOLD - edge_cv) * TEXTURE_ANALYSIS_PARAMS.EDGE_SCALE
# Clipping Sub-anomalies
smoothness_anomaly = np.clip(smoothness_anomaly, 0.0, 1.0)
entropy_anomaly = np.clip(entropy_anomaly, 0.0, 1.0)
contrast_anomaly = np.clip(contrast_anomaly, 0.0, 1.0)
edge_anomaly = np.clip(edge_anomaly, 0.0, 1.0)
# Combine scores
weights = TEXTURE_ANALYSIS_PARAMS.SUBMETRIC_WEIGHTS
texture_score = (weights['smoothness_anomaly'] * smoothness_anomaly + weights['entropy_anomaly'] * entropy_anomaly + weights['contrast_anomaly'] * contrast_anomaly + weights['edge_anomaly'] * edge_anomaly)
final_score = float(np.clip(texture_score, 0.0, 1.0))
detailed_stats = {"smooth_ratio" : float(smooth_ratio),
"entropy_mean" : float(np.mean(entropy_vals)),
"entropy_cv" : float(entropy_cv),
"contrast_mean" : float(np.mean(contrast)),
"contrast_cv" : float(contrast_cv),
"edge_density_mean" : float(np.mean(edge_density)),
"edge_cv" : float(edge_cv),
}
logger.debug(f"Texture scores - smoothness: {smoothness_anomaly:.3f}, entropy: {entropy_anomaly:.3f}, contrast: {contrast_anomaly:.3f}, edge: {edge_anomaly:.3f}")
return final_score, detailed_stats |