Image Augmentation

Introduction

Data augmentation increases training data through random transformations to improve model generalization.

torchvision transforms

from torchvision import transforms
from PIL import Image

transform = transforms.Compose([
    transforms.RandomHorizontalFlip(p=0.5),
    transforms.RandomRotation(degrees=15),
    transforms.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2),
    transforms.RandomResizedCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

img = Image.open('photo.jpg')
augmented = transform(img)

TensorFlow Augmentation

from tensorflow.keras.layers import RandomFlip, RandomRotation, RandomZoom
from tensorflow.keras import layers

data_augmentation = keras.Sequential([
    RandomFlip("horizontal"),
    RandomRotation(0.1),
    RandomZoom(0.1),
])

albumentations

import albumentations as A
import cv2

transform = A.Compose([
    A.HorizontalFlip(p=0.5),
    A.RandomBrightnessContrast(p=0.3),
    A.ShiftScaleRotate(shift_limit=0.1, scale_limit=0.1, rotate_limit=15),
    A.GaussNoise(var_limit=(10.0, 50.0)),
    A.CoarseDropout(max_holes=8, max_height=32, max_width=32)
])

image = cv2.imread('photo.jpg')
augmented = transform(image=image)

Keras ImageDataGenerator

from tensorflow.keras.preprocessing.image import ImageDataGenerator

datagen = ImageDataGenerator(
    rotation_range=20,
    width_shift_range=0.2,
    height_shift_range=0.2,
    horizontal_flip=True,
    zoom_range=0.2,
    fill_mode='nearest'
)

datagen.fit(x_train)

Practice Problems

1. Apply random flips and rotations
2. Add color jitter
3. Implement random cropping
4. Use augmentation in training
5. Create custom augmentation