Skip to main content

An image augmentation library for object detection and image classification tasks.

Project description

augment-auto

A python image augmentation library based on opencv and numpy. It can be used for augmenting images in both image classification and object detection tasks. Many different techniques of augmentation are supported, which can be clustered into three major types - geometric transformations, photometric transformations and kernel-based transformations. Library has support for images with bounding boxes as well.

Installation

Install using pip:

pip install augment-auto

Install from github:

git clone https://github.com/keshavoct98/image-augmentation.git
python setup.py install

Documentation

Complete documentaion - https://augment-auto.readthedocs.io/en/latest/
Demo ipython notebokk - demo.ipynb

Examples

# Geometric Transformations
img = cv2.imread('images/3.jpg')
img_new = crop(img, point1 = (100, 100), point2 = (450, 400))
img_new = rotate(img, angle = 15, keep_resolution = True)
img_new = scale(img, fx = 1.5, fy = 1.5, keep_resolution = False)
img_new = shear(img, shear_val = 0.2, axis = 1)
img_new = translate(img, tx = 50, ty = 60)
# Geometric Transformations with bounding box
img = cv2.imread('images/0.jpeg')
bbox = [581, 274, 699, 321]
img_new, bbox_new = crop(img, point1 = (100, 100), point2 = (650, 400), box = bbox)
img_new, bbox_new = rotate(img, angle = 15, keep_resolution = True, box = bbox)
img_new, bbox_new = scale(img, fx = 1.5, fy = 1.3, keep_resolution = False, box = bbox)
img_new, bbox_new = shear(img, shear_val = 0.2, axis = 0, box = bbox)
img_new, bbox_new = translate(img, tx = 50, ty = 160, box = bbox)
# Photometric Transformations
img = cv2.imread('images/1.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_new = brightness_contrast(img, alpha = 1.3, beta = 20)            
img_new = brightness_contrast(img, alpha = 0.7, beta = -10)
img_new = colorSpace(img, colorspace = 'hsv')             
img_new = colorSpace(img, colorspace = 'ycrcb')           
img_new = colorSpace(img, colorspace = 'lab')
img_new = addNoise(img, 'gaussian', mean = 0, var = 0.08)
img_new = addNoise(img, 'salt_pepper', sp_ratio = 0.5, noise_amount = 0.1)
img_new = addNoise(img, 'poisson', noise_amount = 0.5)
# Kernel-based Transformations
img = cv2.imread('images/0.jpeg')
bbox = [581, 274, 699, 321]
img_new = randomErase(img, size = (100, 100))            
img_new = randomCropAdd(img, size = (100, 100))
img_new = sharpen(img)
img_new = randomErase(img, size = (60, 40), box = bbox)            
img_new = randomCropAdd(img, size = (60, 40), box = bbox)
img_new = blur(img, 'avg', ksize = (9,9))
img_new = blur(img, 'gaussian', ksize = (9,9), gaussian_sigma = 0)
img_new = blur(img, 'median', median_ksize = 11)

References

  1. https://numpy.org/doc/
  2. https://docs.opencv.org/master/
  3. https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_tutorials.html
  4. https://stackabuse.com/affine-image-transformations-in-python-with-numpy-pillow-and-opencv/
  5. https://cristianpb.github.io/blog/image-rotation-opencv

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for augment-auto, version 0.1.0
Filename, size File type Python version Upload date Hashes
Filename, size augment_auto-0.1.0-py3-none-any.whl (8.0 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size augment_auto-0.1.0.tar.gz (6.9 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page