File size: 19,238 Bytes
83e35a7 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 |
"""
Advanced Image Enhancement using State-of-the-Art AI Models
Real-ESRGAN, GFPGAN, and other cutting-edge models
Optimized for NVIDIA RTX 3050
"""
import cv2
import numpy as np
import torch
import torch.nn as nn
from PIL import Image, ImageEnhance, ImageFilter
import os
import requests
from io import BytesIO
import time
from typing import Optional, Tuple
try:
from backend.ai_model_manager import get_ai_model_manager
AI_MODELS_AVAILABLE = True
except ImportError:
AI_MODELS_AVAILABLE = False
print("β οΈ AI models not available, using lightweight enhancer")
from backend.lightweight_ai_enhancer import get_lightweight_enhancer
from backend.compact_ai_models import CompactAIEnhancer
from backend.ultra_compact_enhancer import get_memory_safe_enhancer
class AdvancedImageEnhancer:
"""Advanced image enhancement using state-of-the-art AI models"""
def __init__(self):
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f"π― Using device: {self.device}")
# Check VRAM and decide which enhancer to use
self.use_lightweight = True
if self.device.type == 'cuda':
props = torch.cuda.get_device_properties(0)
vram_gb = props.total_memory / (1024**3)
print(f"π VRAM: {vram_gb:.1f} GB")
# Use lightweight for <6GB VRAM or if heavy models not available
if vram_gb < 6 or not AI_MODELS_AVAILABLE:
self.use_lightweight = True
print("π Using lightweight enhancer (optimized for <4GB VRAM)")
else:
self.use_lightweight = False
# Initialize appropriate manager
if self.use_lightweight:
# Use memory-safe enhancer for <6GB VRAM
print("π Using memory-safe AI enhancer (<1GB VRAM)")
self.enhancer = get_memory_safe_enhancer()
self.ai_manager = None
self.compact_realesrgan = None
else:
self.ai_manager = get_ai_model_manager()
self.enhancer = None
self.compact_realesrgan = None
# Enhancement settings
self.use_ai_models = os.getenv('USE_AI_MODELS', '1') == '1'
self.enhance_faces = os.getenv('ENHANCE_FACES', '1') == '1'
self.use_anime_model = False # Will be set based on content
# Initialize models
self._load_models()
def _load_models(self):
"""Load AI enhancement models"""
try:
if self.use_lightweight:
print("π Loading lightweight AI models...")
# Lightweight models load on demand
self.advanced_available = True
print("β
Lightweight enhancer ready")
else:
print("π Loading advanced AI models...")
if self.use_ai_models and self.ai_manager:
# Load Real-ESRGAN for super resolution
self.ai_manager.load_realesrgan('RealESRGAN_x4plus')
# Pre-load anime model for comic style
self.ai_manager.load_realesrgan('RealESRGAN_x4plus_anime_6B')
# Load GFPGAN for face enhancement
if self.enhance_faces:
self.ai_manager.load_gfpgan()
self.advanced_available = True
print("β
AI models loaded successfully")
else:
print("β οΈ AI models disabled, using traditional methods")
self.advanced_available = False
except Exception as e:
print(f"β οΈ Models failed to load: {e}")
print("β οΈ Falling back to traditional enhancement methods")
self.advanced_available = False
def enhance_image(self, image_path: str, output_path: str = None) -> str:
"""Apply advanced image enhancement"""
if output_path is None:
output_path = image_path
print(f"π Enhancing image: {os.path.basename(image_path)}")
try:
# Load image
img = cv2.imread(image_path)
if img is None:
print(f"β Failed to load image: {image_path}")
return image_path
# Apply enhancement pipeline - pass image_path for compact models
enhanced_img = self._apply_enhancement_pipeline(img, image_path)
# Save enhanced image with maximum quality
cv2.imwrite(output_path, enhanced_img, [cv2.IMWRITE_JPEG_QUALITY, 100])
print(f"β
Enhanced image saved: {os.path.basename(output_path)}")
return output_path
except Exception as e:
print(f"β Enhancement failed: {e}")
return image_path
def _apply_enhancement_pipeline(self, img: np.ndarray, image_path: str = None) -> np.ndarray:
"""Apply complete enhancement pipeline with AI models"""
original_img = img.copy()
print("π¨ Applying AI-powered enhancement pipeline...")
# Detect if image is anime/comic style
self.use_anime_model = self._detect_anime_style(img)
if self.advanced_available and self.use_ai_models:
try:
if self.use_lightweight:
# Use memory-safe enhancer for <4GB VRAM
print(" π Applying memory-safe AI enhancement...")
# Save current image temporarily
temp_path = image_path.replace('.', '_temp.')
cv2.imwrite(temp_path, img)
# Process with memory-safe enhancer
enhanced_path = self.enhancer.enhance_image(
temp_path,
temp_path.replace('_temp.', '_enhanced.')
)
# Read enhanced image
img = cv2.imread(enhanced_path)
# Clean up temp files
if os.path.exists(temp_path):
os.remove(temp_path)
if os.path.exists(enhanced_path) and enhanced_path != image_path:
os.remove(enhanced_path)
print(" β
Memory-safe enhancement complete")
# Show memory usage
if hasattr(self.enhancer, 'get_memory_usage'):
print(f" πΎ Memory: {self.enhancer.get_memory_usage()}")
else:
# Use full AI models for >6GB VRAM
print(" π Applying AI super resolution...")
img = self.ai_manager.enhance_image_realesrgan(
img,
use_anime_model=self.use_anime_model
)
# 2. AI Face Enhancement with GFPGAN
if self.enhance_faces:
print(" π€ Enhancing faces with AI...")
img = self.ai_manager.enhance_face_gfpgan(img)
# 3. Post-processing
img = self.ai_manager.post_process(img)
# Clear GPU memory
self.ai_manager.clear_memory()
return img
except Exception as e:
print(f"β οΈ AI enhancement failed: {e}, using fallback")
img = original_img
# Fallback to traditional methods if AI models not available
print(" π Using traditional enhancement methods...")
# 1. Traditional Super Resolution
img = self._apply_super_resolution_advanced(img)
# 2. Advanced Color Enhancement
img = self._enhance_colors_advanced(img)
# 3. Advanced Noise Reduction
img = self._reduce_noise_advanced(img)
# 4. Advanced Sharpness Enhancement
img = self._enhance_sharpness_advanced(img)
# 5. Advanced Dynamic Range Optimization
img = self._optimize_dynamic_range_advanced(img)
# 6. Traditional Face Enhancement
img = self._enhance_faces_advanced(img)
return img
def _apply_super_resolution_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced super resolution (4x upscaling)"""
try:
print("π Applying advanced super resolution (4x upscaling)...")
# Get original dimensions
height, width = img.shape[:2]
# Calculate target dimensions (max 2K - 2048x1080)
scale_factor = min(2048 / width, 1080 / height, 2.0) # Max 2x upscaling
target_width = int(width * scale_factor)
target_height = int(height * scale_factor)
# Use LANCZOS interpolation for highest quality
img = cv2.resize(img, (target_width, target_height),
interpolation=cv2.INTER_LANCZOS4)
# Apply additional sharpening after upscaling
kernel = np.array([[-1,-1,-1], [-1,9,-1], [-1,-1,-1]])
img = cv2.filter2D(img, -1, kernel)
print(f"β
Super resolution completed: {width}x{height} β {target_width}x{target_height}")
except Exception as e:
print(f"β οΈ Super resolution failed: {e}")
return img
def _enhance_colors_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced color enhancement"""
try:
print("π¨ Applying advanced color enhancement...")
# Convert to LAB color space for better color processing
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
# Enhance L channel (lightness) with CLAHE
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8,8))
lab[:,:,0] = clahe.apply(lab[:,:,0])
# Enhance A and B channels (color) with adaptive scaling
lab[:,:,1] = cv2.convertScaleAbs(lab[:,:,1], alpha=1.3, beta=10)
lab[:,:,2] = cv2.convertScaleAbs(lab[:,:,2], alpha=1.3, beta=10)
# Convert back to BGR
enhanced = cv2.cvtColor(lab, cv2.COLOR_LAB2BGR)
# Additional color saturation enhancement
hsv = cv2.cvtColor(enhanced, cv2.COLOR_BGR2HSV)
hsv[:,:,1] = cv2.convertScaleAbs(hsv[:,:,1], alpha=1.4, beta=0) # Increase saturation
enhanced = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
except Exception as e:
print(f"β οΈ Color enhancement failed: {e}")
enhanced = img
return enhanced
def _reduce_noise_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced noise reduction"""
try:
print("π§Ή Applying advanced noise reduction...")
# Multi-stage noise reduction
# 1. Bilateral filter for edge-preserving smoothing
denoised = cv2.bilateralFilter(img, 9, 75, 75)
# 2. Non-local means denoising for additional noise reduction
denoised = cv2.fastNlMeansDenoisingColored(denoised, None, 10, 10, 7, 21)
# 3. Gaussian blur for final smoothing
denoised = cv2.GaussianBlur(denoised, (3, 3), 0)
# 4. Edge-preserving filter
denoised = cv2.edgePreservingFilter(denoised, flags=1, sigma_s=60, sigma_r=0.4)
except Exception as e:
print(f"β οΈ Noise reduction failed: {e}")
denoised = img
return denoised
def _enhance_sharpness_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced sharpness enhancement"""
try:
print("πͺ Applying advanced sharpness enhancement...")
# Multi-stage sharpening
# 1. Unsharp masking
gaussian = cv2.GaussianBlur(img, (0, 0), 2.0)
sharpened = cv2.addWeighted(img, 1.5, gaussian, -0.5, 0)
# 2. Edge enhancement
kernel = np.array([[-1,-1,-1], [-1,9,-1], [-1,-1,-1]])
sharpened = cv2.filter2D(sharpened, -1, kernel)
# 3. Laplacian sharpening
gray = cv2.cvtColor(sharpened, cv2.COLOR_BGR2GRAY)
laplacian = cv2.Laplacian(gray, cv2.CV_64F)
laplacian = np.uint8(np.absolute(laplacian))
sharpened = cv2.addWeighted(sharpened, 1.0, cv2.cvtColor(laplacian, cv2.COLOR_GRAY2BGR), 0.3, 0)
except Exception as e:
print(f"β οΈ Sharpness enhancement failed: {e}")
sharpened = img
return sharpened
def _optimize_dynamic_range_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced dynamic range optimization"""
try:
print("π Applying advanced dynamic range optimization...")
# Convert to LAB color space
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
# Apply CLAHE to L channel for better contrast
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
lab[:,:,0] = clahe.apply(lab[:,:,0])
# Enhance contrast in A and B channels
lab[:,:,1] = cv2.convertScaleAbs(lab[:,:,1], alpha=1.2, beta=0)
lab[:,:,2] = cv2.convertScaleAbs(lab[:,:,2], alpha=1.2, beta=0)
# Convert back to BGR
optimized = cv2.cvtColor(lab, cv2.COLOR_LAB2BGR)
# Additional contrast enhancement
optimized = cv2.convertScaleAbs(optimized, alpha=1.1, beta=5)
except Exception as e:
print(f"β οΈ Dynamic range optimization failed: {e}")
optimized = img
return optimized
def _enhance_faces_advanced(self, img: np.ndarray) -> np.ndarray:
"""Advanced face enhancement"""
try:
print("π€ Applying advanced face enhancement...")
# Load face cascade
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
# Detect faces
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, 1.1, 4)
if len(faces) > 0:
print(f"π Found {len(faces)} faces, applying enhancement...")
for (x, y, w, h) in faces:
# Extract face region
face_roi = img[y:y+h, x:x+w]
# Apply face-specific enhancement
enhanced_face = self._enhance_face_region(face_roi)
# Replace face region
img[y:y+h, x:x+w] = enhanced_face
else:
print("π€ No faces detected, skipping face enhancement")
except Exception as e:
print(f"β οΈ Face enhancement failed: {e}")
return img
def _enhance_face_region(self, face_img: np.ndarray) -> np.ndarray:
"""Enhance a specific face region"""
try:
# Apply gentle smoothing to face
enhanced = cv2.bilateralFilter(face_img, 5, 50, 50)
# Enhance skin tone
hsv = cv2.cvtColor(enhanced, cv2.COLOR_BGR2HSV)
hsv[:,:,1] = cv2.convertScaleAbs(hsv[:,:,1], alpha=1.1, beta=0) # Gentle saturation boost
enhanced = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
# Apply subtle sharpening
kernel = np.array([[-0.5,-0.5,-0.5], [-0.5,5,-0.5], [-0.5,-0.5,-0.5]])
enhanced = cv2.filter2D(enhanced, -1, kernel)
except Exception as e:
enhanced = face_img
return enhanced
def _detect_anime_style(self, img: np.ndarray) -> bool:
"""Detect if image is anime/manga/comic style"""
try:
# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 1. Edge density check - anime has cleaner edges
edges = cv2.Canny(gray, 50, 150)
edge_density = np.sum(edges > 0) / edges.size
# 2. Color count check - anime has fewer unique colors
unique_colors = len(np.unique(img.reshape(-1, img.shape[2]), axis=0))
# 3. Gradient smoothness - anime has smoother gradients
laplacian = cv2.Laplacian(gray, cv2.CV_64F)
gradient_variance = np.var(laplacian)
# Decision logic
is_anime = (
edge_density < 0.15 and # Clean edges
unique_colors < 10000 and # Limited color palette
gradient_variance < 1000 # Smooth gradients
)
if is_anime:
print(" π Detected anime/comic style - using specialized model")
return is_anime
except Exception as e:
print(f"β οΈ Style detection failed: {e}")
return False
def enhance_batch(self, image_paths: list, output_dir: str = None) -> list:
"""Enhance multiple images"""
if output_dir is None:
output_dir = "enhanced"
os.makedirs(output_dir, exist_ok=True)
enhanced_paths = []
print(f"π― Enhancing {len(image_paths)} images with advanced techniques...")
for i, image_path in enumerate(image_paths, 1):
print(f"πΈ Processing {i}/{len(image_paths)}: {os.path.basename(image_path)}")
# Generate output path
filename = os.path.basename(image_path)
output_path = os.path.join(output_dir, f"enhanced_{filename}")
# Enhance image
enhanced_path = self.enhance_image(image_path, output_path)
enhanced_paths.append(enhanced_path)
print(f"β
Enhanced {len(enhanced_paths)} images with advanced techniques")
return enhanced_paths
# Global instance
advanced_enhancer = None
def get_advanced_enhancer():
"""Get or create global advanced enhancer instance"""
global advanced_enhancer
if advanced_enhancer is None:
advanced_enhancer = AdvancedImageEnhancer()
return advanced_enhancer
if __name__ == "__main__":
# Test the enhancer
enhancer = AdvancedImageEnhancer()
print("π§ͺ Advanced Image Enhancer ready for testing!") |