File size: 12,031 Bytes
6524e7a |
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 |
# Create an improved contrast detection file: utils/improved_contrast_analyzer.py
import numpy as np
import cv2
import colorsys
from scipy import ndimage, spatial
from sklearn.cluster import DBSCAN
class ImprovedContrastAnalyzer:
"""
Advanced contrast analyzer focused on Alzheimer's-friendly environments
"""
def __init__(self, wcag_threshold=4.5):
self.wcag_threshold = wcag_threshold
# ADE20K class mappings for important objects
self.important_classes = {
'floor': [3, 4], # floor, wood floor
'wall': [0, 1], # wall, building
'ceiling': [5], # ceiling
'sofa': [10], # sofa
'chair': [19], # chair
'table': [15], # table
'door': [25], # door
'window': [8], # window
'stairs': [53], # stairs
'bed': [7], # bed
}
# Priority relationships (high priority = more important for safety)
self.priority_relationships = {
('floor', 'sofa'): 'high',
('floor', 'chair'): 'high',
('floor', 'table'): 'high',
('wall', 'sofa'): 'medium',
('wall', 'chair'): 'medium',
('wall', 'door'): 'high',
('floor', 'stairs'): 'critical',
('floor', 'bed'): 'medium',
('wall', 'window'): 'low',
('ceiling', 'wall'): 'low',
}
def get_object_category(self, class_id):
"""Map segmentation class to object category"""
for category, class_ids in self.important_classes.items():
if class_id in class_ids:
return category
return 'other'
def calculate_wcag_contrast(self, color1, color2):
"""Calculate WCAG contrast ratio"""
def relative_luminance(color):
rgb = [c / 255.0 for c in color]
return sum(c / 12.92 if c <= 0.03928 else ((c + 0.055) / 1.055) ** 2.4
for c in rgb) * [0.2126, 0.7152, 0.0722][i] for i, c in enumerate(rgb))
lum1 = sum(self.relative_luminance_component(color1))
lum2 = sum(self.relative_luminance_component(color2))
lighter = max(lum1, lum2)
darker = min(lum1, lum2)
return (lighter + 0.05) / (darker + 0.05)
def relative_luminance_component(self, color):
"""Calculate relative luminance components"""
rgb = [c / 255.0 for c in color]
components = []
factors = [0.2126, 0.7152, 0.0722]
for i, c in enumerate(rgb):
if c <= 0.03928:
components.append((c / 12.92) * factors[i])
else:
components.append(((c + 0.055) / 1.055) ** 2.4 * factors[i])
return components
def calculate_perceptual_contrast(self, color1, color2):
"""Calculate perceptual contrast including hue and saturation differences"""
# Convert to HSV for better perceptual analysis
hsv1 = cv2.cvtColor(np.uint8([[color1]]), cv2.COLOR_RGB2HSV)[0][0] / 255.0
hsv2 = cv2.cvtColor(np.uint8([[color2]]), cv2.COLOR_RGB2HSV)[0][0] / 255.0
# Hue difference (circular)
hue_diff = abs(hsv1[0] - hsv2[0])
if hue_diff > 0.5:
hue_diff = 1 - hue_diff
# Saturation difference
sat_diff = abs(hsv1[1] - hsv2[1])
# Value (brightness) difference
val_diff = abs(hsv1[2] - hsv2[2])
# Combined perceptual score (0-1, higher is more different)
perceptual_contrast = np.sqrt(hue_diff**2 + sat_diff**2 + val_diff**2) / np.sqrt(3)
return perceptual_contrast
def find_clean_boundaries(self, mask1, mask2, min_boundary_length=50):
"""Find clean boundaries between two segments"""
# Dilate both masks slightly
kernel = np.ones((3, 3), np.uint8)
dilated1 = cv2.dilate(mask1.astype(np.uint8), kernel, iterations=1)
dilated2 = cv2.dilate(mask2.astype(np.uint8), kernel, iterations=1)
# Find intersection (boundary area)
boundary = (dilated1 & dilated2).astype(bool)
# Remove small disconnected boundary pieces
labeled_boundary = ndimage.label(boundary)[0]
for region_id in range(1, labeled_boundary.max() + 1):
region_mask = labeled_boundary == region_id
if np.sum(region_mask) < min_boundary_length:
boundary[region_mask] = False
return boundary
def get_representative_colors(self, image, mask, n_samples=1000):
"""Get representative colors from a masked region using clustering"""
if not np.any(mask):
return np.array([0, 0, 0])
# Sample pixels from the mask
y_coords, x_coords = np.where(mask)
if len(y_coords) > n_samples:
indices = np.random.choice(len(y_coords), n_samples, replace=False)
y_coords = y_coords[indices]
x_coords = x_coords[indices]
colors = image[y_coords, x_coords]
# Use DBSCAN clustering to find dominant colors
if len(colors) > 10:
clustering = DBSCAN(eps=30, min_samples=5).fit(colors)
labels = clustering.labels_
# Get the largest cluster
unique_labels, counts = np.unique(labels[labels >= 0], return_counts=True)
if len(unique_labels) > 0:
dominant_label = unique_labels[np.argmax(counts)]
dominant_colors = colors[labels == dominant_label]
return np.mean(dominant_colors, axis=0).astype(int)
# Fallback to mean color
return np.mean(colors, axis=0).astype(int)
def analyze_improved_contrast(self, image, segmentation):
"""
Perform improved contrast analysis focused on important relationships
"""
h, w = segmentation.shape
results = {
'critical_issues': [],
'high_priority_issues': [],
'medium_priority_issues': [],
'statistics': {},
'visualization': image.copy()
}
# Get unique segments and their categories
unique_segments = np.unique(segmentation)
segment_categories = {}
segment_colors = {}
for seg_id in unique_segments:
if seg_id == 0: # Skip background
continue
mask = segmentation == seg_id
category = self.get_object_category(seg_id)
segment_categories[seg_id] = category
segment_colors[seg_id] = self.get_representative_colors(image, mask)
# Analyze important relationships
total_issues = 0
critical_count = 0
high_count = 0
medium_count = 0
for i, seg_id1 in enumerate(unique_segments):
if seg_id1 == 0:
continue
category1 = segment_categories.get(seg_id1, 'other')
if category1 == 'other':
continue
for seg_id2 in unique_segments[i+1:]:
if seg_id2 == 0:
continue
category2 = segment_categories.get(seg_id2, 'other')
if category2 == 'other':
continue
# Check if this is an important relationship
relationship = tuple(sorted([category1, category2]))
priority = self.priority_relationships.get(relationship)
if priority is None:
continue
# Check if segments are adjacent
mask1 = segmentation == seg_id1
mask2 = segmentation == seg_id2
boundary = self.find_clean_boundaries(mask1, mask2)
if not np.any(boundary):
continue
# Calculate contrasts
color1 = segment_colors[seg_id1]
color2 = segment_colors[seg_id2]
wcag_contrast = self.calculate_wcag_contrast(color1, color2)
perceptual_contrast = self.calculate_perceptual_contrast(color1, color2)
# Determine if there's an issue
wcag_issue = wcag_contrast < self.wcag_threshold
perceptual_issue = perceptual_contrast < 0.3 # Threshold for perceptual difference
if wcag_issue or perceptual_issue:
issue = {
'categories': (category1, category2),
'segment_ids': (seg_id1, seg_id2),
'wcag_contrast': wcag_contrast,
'perceptual_contrast': perceptual_contrast,
'boundary_pixels': np.sum(boundary),
'priority': priority
}
# Color-code the boundary based on priority
if priority == 'critical':
results['critical_issues'].append(issue)
results['visualization'][boundary] = [255, 0, 0] # Red
critical_count += 1
elif priority == 'high':
results['high_priority_issues'].append(issue)
results['visualization'][boundary] = [255, 128, 0] # Orange
high_count += 1
elif priority == 'medium':
results['medium_priority_issues'].append(issue)
results['visualization'][boundary] = [255, 255, 0] # Yellow
medium_count += 1
total_issues += 1
# Calculate statistics
results['statistics'] = {
'total_issues': total_issues,
'critical_issues': critical_count,
'high_priority_issues': high_count,
'medium_priority_issues': medium_count,
'segments_analyzed': len([cat for cat in segment_categories.values() if cat != 'other'])
}
return results
# Update your contrast detection imports and usage
class PrioritizedContrastDetector:
"""Wrapper for the improved contrast analyzer"""
def __init__(self, threshold=4.5):
self.analyzer = ImprovedContrastAnalyzer(wcag_threshold=threshold)
def analyze(self, image, segmentation, threshold, highlight_color=(255, 0, 0)):
"""Analyze with improved logic"""
results = self.analyzer.analyze_improved_contrast(image, segmentation)
# Convert to format expected by original interface
contrast_image = results['visualization']
# Create a simple problem areas mask for compatibility
problem_areas = np.any([
contrast_image[:, :, 0] == 255, # Any red channel highlighting
], axis=0)
# Format statistics
stats = results['statistics'].copy()
stats['threshold'] = threshold
stats['problem_count'] = stats['total_issues']
# Add detailed breakdown
if results['critical_issues']:
stats['critical_details'] = [
f"{issue['categories'][0]}-{issue['categories'][1]}: WCAG {issue['wcag_contrast']:.1f}:1"
for issue in results['critical_issues']
]
if results['high_priority_issues']:
stats['high_priority_details'] = [
f"{issue['categories'][0]}-{issue['categories'][1]}: WCAG {issue['wcag_contrast']:.1f}:1"
for issue in results['high_priority_issues']
]
return contrast_image, problem_areas, stats
|