Datasets:
DOI:
License:
File size: 10,314 Bytes
5ed72c3 bb93c50 404e2b5 bb93c50 404e2b5 bb93c50 | 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 | # Copyright 2025 Robotics Group of the University of León (ULE)
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import random
from pathlib import Path
import numpy as np
from PIL import Image, ImageEnhance, ImageOps, ImageFilter
from tqdm import tqdm
def load_image(image_path):
"""Load an image from the file system.
Args:
image_path: Path to the image file.
Returns:
Loaded PIL Image object.
"""
return Image.open(image_path)
def save_image(image, output_path):
"""Save the image to the specified path.
Args:
image: PIL Image object to save.
output_path: Destination path for the image.
"""
image.save(output_path)
def flip_horizontal(image):
"""Flip the image horizontally.
Args:
image: Input PIL Image.
Returns:
Horizontally flipped PIL Image.
"""
return ImageOps.mirror(image)
def rotate(image, angle):
"""Rotate the image by the specified angle.
Args:
image: Input PIL Image.
angle: Rotation angle in degrees (positive = counter-clockwise).
Returns:
Rotated PIL Image.
"""
return image.rotate(angle, expand=False)
def adjust_brightness(image, factor):
"""Adjust the brightness of the image.
Args:
image: Input PIL Image.
factor: Brightness adjustment factor (1.0 = original, <1.0 darker, >1.0 brighter).
Returns:
Brightness-adjusted PIL Image.
"""
enhancer = ImageEnhance.Brightness(image)
return enhancer.enhance(factor)
def adjust_contrast(image, factor):
"""Adjust the contrast of the image.
Args:
image: Input PIL Image.
factor: Contrast adjustment factor (1.0 = original, <1.0 less contrast, >1.0 more contrast).
Returns:
Contrast-adjusted PIL Image.
"""
enhancer = ImageEnhance.Contrast(image)
return enhancer.enhance(factor)
def adjust_saturation(image, factor):
"""Adjust the saturation of the image.
Args:
image: Input PIL Image.
factor: Saturation adjustment factor (1.0 = original, 0.0 = grayscale, >1.0 more saturated).
Returns:
Saturation-adjusted PIL Image.
"""
enhancer = ImageEnhance.Color(image)
return enhancer.enhance(factor)
def add_noise(image, intensity=0.05):
"""Add random noise to the image.
Args:
image: Input PIL Image.
intensity: Noise intensity factor (0.0-1.0). Defaults to 0.05.
Returns:
PIL Image with added noise.
"""
img_array = np.array(image).copy()
if len(img_array.shape) == 3:
h, w, c = img_array.shape
noise = np.random.randint(-int(intensity * 255), int(intensity * 255), (h, w, c))
img_array = np.clip(img_array + noise, 0, 255).astype(np.uint8)
else:
h, w = img_array.shape
noise = np.random.randint(-int(intensity * 255), int(intensity * 255), (h, w))
img_array = np.clip(img_array + noise, 0, 255).astype(np.uint8)
return Image.fromarray(img_array)
def apply_blur(image, radius=2):
"""Apply Gaussian blur to the image.
Args:
image: Input PIL Image.
radius: Blur radius in pixels. Defaults to 2.
Returns:
Blurred PIL Image.
"""
return image.filter(ImageFilter.GaussianBlur(radius=radius))
def transform_yolo_label(label_line, technique_name):
"""Transform YOLO format label coordinates based on the augmentation technique.
Args:
label_line: A line from a YOLO label file (class_id center_x center_y width height).
technique_name: The augmentation technique applied.
Returns:
Transformed label line in YOLO format.
"""
parts = label_line.strip().split()
if len(parts) < 5:
return label_line
class_id = parts[0]
x_center = float(parts[1])
y_center = float(parts[2])
width = float(parts[3])
height = float(parts[4])
if technique_name == "flip_horizontal":
x_center = 1.0 - x_center
return f"{class_id} {x_center:.6f} {y_center:.6f} {width:.6f} {height:.6f}"
def process_yolo_label(original_label_path, new_label_path, technique_name):
"""Process a YOLO format label file, applying transformations to the coordinates.
Args:
original_label_path: Path to the original label file.
new_label_path: Path for the new transformed label file.
technique_name: Name of the augmentation technique applied.
Returns:
True if processing was successful, False otherwise.
"""
if not original_label_path.exists():
return False
with open(original_label_path, "r") as infile:
lines = infile.readlines()
new_lines = []
for line in lines:
if line.strip():
if technique_name in ["adjust_brightness", "adjust_contrast",
"adjust_saturation", "add_noise", "apply_blur"]:
new_lines.append(line.strip())
else:
transformed_line = transform_yolo_label(line, technique_name)
new_lines.append(transformed_line)
os.makedirs(new_label_path.parent, exist_ok=True)
with open(new_label_path, "w") as outfile:
outfile.write("\n".join(new_lines))
return True
def augment_yolo_dataset(base_dir, augmentations_per_image=3):
"""Apply data augmentation to images in a YOLO dataset.
For each image in the dataset, this function applies a specified number of random
augmentation techniques and transforms the corresponding YOLO label files to maintain
annotation accuracy.
Args:
base_dir: Base directory of the YOLO dataset (should contain images/ and labels/).
augmentations_per_image: Number of random augmentations per image. Defaults to 3.
"""
images_dir = os.path.join(base_dir, "images")
labels_dir = os.path.join(base_dir, "labels")
if not os.path.exists(images_dir):
print(f"Error: Images directory not found at {images_dir}")
return
if not os.path.exists(labels_dir):
print(f"Error: Labels directory not found at {labels_dir}")
return
image_extensions = [".jpg", ".jpeg", ".png", ".bmp"]
image_files = []
for ext in image_extensions:
image_files.extend(list(Path(images_dir).glob(f"*{ext}")))
print(f"Found {len(image_files)} images in {images_dir}")
augmentation_techniques = [
(flip_horizontal, {}, "flip_horizontal"),
(rotate, {"angle": lambda: random.randint(-15, 15)}, "rotate"),
(adjust_brightness, {"factor": lambda: random.uniform(0.8, 1.2)}, "adjust_brightness"),
(adjust_contrast, {"factor": lambda: random.uniform(0.8, 1.2)}, "adjust_contrast"),
(adjust_saturation, {"factor": lambda: random.uniform(0.8, 1.2)}, "adjust_saturation"),
(add_noise, {"intensity": lambda: random.uniform(0.01, 0.05)}, "add_noise"),
(apply_blur, {"radius": lambda: random.uniform(0.5, 1.5)}, "apply_blur")
]
total_augmented = 0
labels_processed = 0
for img_path in tqdm(image_files, desc="Augmenting images"):
try:
original_image = load_image(img_path)
rel_path = img_path.relative_to(images_dir)
label_path = Path(labels_dir) / rel_path.with_suffix(".txt")
if not label_path.exists():
continue
for i in range(augmentations_per_image):
technique, params, technique_name = random.choice(augmentation_techniques)
resolved_params = {k: v() if callable(v) else v for k, v in params.items()}
augmented_image = original_image.copy()
augmented_image = technique(augmented_image, **resolved_params)
filename = img_path.stem
extension = img_path.suffix
augmented_filename = f"{filename}_aug_{technique_name}_{i + 1}{extension}"
augmented_path = img_path.parent / augmented_filename
save_image(augmented_image, augmented_path)
augmented_label_path = label_path.parent / f"{filename}_aug_{technique_name}_{i + 1}.txt"
if process_yolo_label(label_path, augmented_label_path, technique_name):
labels_processed += 1
total_augmented += 1
except Exception as e:
tqdm.write(f"Error processing {img_path.name}: {str(e)}")
continue
print(f"Created {total_augmented} augmented images")
print(f"Processed {labels_processed} YOLO label files")
train_txt_path = os.path.join(base_dir, "train.txt")
if os.path.exists(train_txt_path):
all_images = []
for ext in image_extensions:
all_images.extend([str(p.relative_to(base_dir)) for p in Path(images_dir).glob(f"*{ext}")])
with open(train_txt_path, "w") as f:
f.write("\n".join(all_images))
print(f"Updated {train_txt_path} with {len(all_images)} image paths")
def main():
"""Main function to execute dataset augmentation.
Configures the base directory and number of augmentations, then runs the
augmentation pipeline on the YOLO dataset.
"""
train_dir = "/home/pcrn/datainbrief/beet_augmented/train"
augmentations_per_image = 3
print(f"Starting augmentation on YOLO dataset in {train_dir}")
augment_yolo_dataset(train_dir, augmentations_per_image)
if __name__ == "__main__":
main() |