File size: 10,562 Bytes
5b86813 | 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 | #!/usr/bin/env python3
"""
Convert YOLO-format object detection dataset to GAN-ready patch dataset.
This script reads YOLO annotations and extracts object patches from images,
resizing them to 64×64 for use in GAN training.
Structure:
Input: dataset/train(valid)/{images,labels}
Output: gan_dataset/{pothole,cracks,open_manhole}/
"""
import os
import sys
from pathlib import Path
import cv2
# Dataset configuration
DATASET_ROOT = Path("/home/pragadeesh/ARM/model/dataset")
OUTPUT_ROOT = Path("/home/pragadeesh/ARM/model/gan_dataset")
# Class mapping from YOLO format
CLASS_NAMES = {
0: "pothole",
1: "cracks",
2: "open_manhole",
3: "good_road"
}
# Classes to ignore (good_road)
IGNORE_CLASSES = {3}
# Target patch size for GAN training
PATCH_SIZE = 64
# Minimum object size in pixels (skip smaller objects)
MIN_WIDTH = 10
MIN_HEIGHT = 10
def setup_output_directories():
"""Create output directory structure for each class."""
OUTPUT_ROOT.mkdir(parents=True, exist_ok=True)
for class_id, class_name in CLASS_NAMES.items():
if class_id not in IGNORE_CLASSES:
class_dir = OUTPUT_ROOT / class_name
class_dir.mkdir(parents=True, exist_ok=True)
print(f"✓ Created: {class_dir}")
def parse_yolo_label(label_path, img_width, img_height):
"""
Parse YOLO format label file.
YOLO format: class_id x_center y_center width height
All values normalized to [0, 1]
Args:
label_path: Path to .txt label file
img_width: Image width in pixels
img_height: Image height in pixels
Returns:
List of tuples: (class_id, x1, y1, x2, y2) in pixel coordinates
"""
bboxes = []
# Return empty list if label file doesn't exist
if not label_path.exists():
return bboxes
try:
with open(label_path, 'r') as f:
for line_num, line in enumerate(f, 1):
line = line.strip()
# Skip empty lines
if not line:
continue
try:
parts = line.split()
# Validate format
if len(parts) < 5:
continue
# Parse YOLO annotation
class_id = int(parts[0])
x_center_norm = float(parts[1])
y_center_norm = float(parts[2])
width_norm = float(parts[3])
height_norm = float(parts[4])
# Convert normalized coordinates to pixel coordinates
x_center_px = x_center_norm * img_width
y_center_px = y_center_norm * img_height
width_px = width_norm * img_width
height_px = height_norm * img_height
# Calculate bounding box corners (top-left and bottom-right)
x1 = max(0, int(x_center_px - width_px / 2))
y1 = max(0, int(y_center_px - height_px / 2))
x2 = min(img_width, int(x_center_px + width_px / 2))
y2 = min(img_height, int(y_center_px + height_px / 2))
# Add to list
bboxes.append((class_id, x1, y1, x2, y2))
except (ValueError, IndexError) as e:
# Gracefully skip malformed lines
pass
except Exception as e:
# Log but don't crash on read errors
pass
return bboxes
def extract_patch(image, bbox, class_name):
"""
Extract a patch from image and resize to 64×64.
Args:
image: OpenCV image (BGR)
bbox: Tuple (class_id, x1, y1, x2, y2) in pixel coordinates
class_name: Name of class for validation
Returns:
Resized patch (64×64) or None if invalid
"""
class_id, x1, y1, x2, y2 = bbox
# Validate bounding box coordinates
if x1 >= x2 or y1 >= y2:
return None
patch_width = x2 - x1
patch_height = y2 - y1
# Skip patches that are too small
if patch_width < MIN_WIDTH or patch_height < MIN_HEIGHT:
return None
# Extract region from image
try:
patch = image[y1:y2, x1:x2]
# Validate patch
if patch.size == 0:
return None
# Resize to 64×64
patch_resized = cv2.resize(patch, (PATCH_SIZE, PATCH_SIZE))
return patch_resized
except Exception as e:
# Handle any extraction errors
return None
def process_dataset_split(split_name):
"""
Process all images in a dataset split (train or valid).
Args:
split_name: String 'train' or 'valid'
Returns:
Dictionary with processing statistics
"""
stats = {
'images_processed': 0,
'labels_read': 0,
'patches_extracted': 0,
'patches_saved': 0,
'errors': 0,
'by_class': {}
}
# Construct paths
split_root = DATASET_ROOT / split_name
images_dir = split_root / 'images'
labels_dir = split_root / 'labels'
# Validate directories exist
if not images_dir.exists():
print(f"✗ Images directory not found: {images_dir}")
return stats
print(f"\n{'='*70}")
print(f"Processing {split_name.upper()} split")
print(f"{'='*70}")
print(f"Images dir: {images_dir}")
print(f"Labels dir: {labels_dir}")
# Find all image files
image_extensions = ('.jpg', '.jpeg', '.png', '.JPG', '.JPEG', '.PNG')
image_paths = sorted([
p for p in images_dir.iterdir()
if p.suffix.lower() in [e.lower() for e in image_extensions]
])
print(f"Found {len(image_paths)} images")
if not image_paths:
print(f"⚠ No images found in {images_dir}")
return stats
# Counter for patch filenames per class
patch_counters = {
'pothole': 0,
'cracks': 0,
'open_manhole': 0
}
# Process each image
for img_idx, image_path in enumerate(image_paths, 1):
# Corresponding label file
label_path = labels_dir / (image_path.stem + '.txt')
# Read image
try:
image = cv2.imread(str(image_path))
if image is None:
stats['errors'] += 1
continue
except Exception as e:
stats['errors'] += 1
continue
img_height, img_width = image.shape[:2]
# Parse YOLO labels
bboxes = parse_yolo_label(label_path, img_width, img_height)
if bboxes:
stats['labels_read'] += 1
# Process each bounding box
for bbox in bboxes:
class_id = bbox[0]
# Skip ignored classes (good_road)
if class_id in IGNORE_CLASSES:
continue
# Get class name
if class_id not in CLASS_NAMES:
continue
class_name = CLASS_NAMES[class_id]
# Extract and resize patch
patch = extract_patch(image, bbox, class_name)
if patch is None:
continue
stats['patches_extracted'] += 1
# Save patch
output_dir = OUTPUT_ROOT / class_name
patch_counters[class_name] += 1
output_filename = f"{class_name}_{patch_counters[class_name]:06d}.jpg"
output_path = output_dir / output_filename
try:
success = cv2.imwrite(str(output_path), patch)
if success:
stats['patches_saved'] += 1
stats['by_class'][class_name] = stats['by_class'].get(class_name, 0) + 1
else:
stats['errors'] += 1
except Exception as e:
stats['errors'] += 1
stats['images_processed'] += 1
# Progress indicator every 50 images
if img_idx % 50 == 0:
print(f" Progress: {img_idx}/{len(image_paths)} images")
return stats
def main():
"""Main entry point."""
print("\n" + "="*70)
print("YOLO to GAN Dataset Converter")
print("="*70)
print(f"Dataset root: {DATASET_ROOT}")
print(f"Output root: {OUTPUT_ROOT}")
print(f"Patch size: {PATCH_SIZE}×{PATCH_SIZE}")
print(f"Min dimensions: {MIN_WIDTH}×{MIN_HEIGHT}")
# Check dataset exists
if not DATASET_ROOT.exists():
print(f"\n✗ Dataset directory not found: {DATASET_ROOT}")
sys.exit(1)
# Setup output structure
print("\nSetting up output directories...")
setup_output_directories()
# Process both splits
all_stats = {}
for split in ['train', 'valid']:
stats = process_dataset_split(split)
all_stats[split] = stats
# Print summary report
print("\n" + "="*70)
print("CONVERSION COMPLETE - SUMMARY")
print("="*70)
total_images = 0
total_patches = 0
for split in ['train', 'valid']:
stats = all_stats[split]
print(f"\n{split.upper()} SET:")
print(f" Images processed: {stats['images_processed']}")
print(f" Labels found: {stats['labels_read']}")
print(f" Patches extracted: {stats['patches_extracted']}")
print(f" Patches saved: {stats['patches_saved']}")
if stats['errors'] > 0:
print(f" Errors: {stats['errors']}")
if stats['by_class']:
print(f" By class:")
for class_name in sorted(stats['by_class'].keys()):
count = stats['by_class'][class_name]
print(f" • {class_name}: {count}")
total_images += stats['images_processed']
total_patches += stats['patches_saved']
print("\n" + "-"*70)
print(f"TOTAL IMAGES: {total_images}")
print(f"TOTAL PATCHES: {total_patches}")
print(f"OUTPUT: {OUTPUT_ROOT}")
print("="*70 + "\n")
if total_patches == 0:
print("⚠ No patches extracted. Check dataset structure and labels.")
sys.exit(1)
print("✓ Done!\n")
if __name__ == "__main__":
main() |