Smallest max size

If you have ever worked on a Computer Vision project, you might know that using augmentations to diversify the dataset is the best practice. On this page, we will:

• Сover the Smallest max size augmentation;

• Check out its parameters;

• See how Smallest max size affects an image;

• And check out how to work with Smallest max size using Python through the Albumentations library.

Let's jump in.

Smallest max size explained

To define the term, Smallest max size is a data augmentation technique that reshapes an image so that the length of an image’s smallest size is equal to a certain number of pixels.

Parameters

• Maximum size of smallest side - sets the desired maximal length of the image in pixels. The values can vary from 256 to 3000.

Advanced options

• Interpolation - used to specify the interpolation algorithm;
  • Nearest;
  • Linear;
  • Cubic;
  • Area;
  • LANCZOS4.

Smallest max Size visualized

Code Implementation

    import albumentations as albu
from PIL import Image
import numpy as np

transform = albu.SmallestMaxSize(max_size=1024, interpolation=1, p=1)
#default interpolation is INTER_LINEAR

image = np.array(Image.open('/some/image/file/path'))
image = transform(image=image)['image']

# Now the image is preprocessed and ready to be accepted by the model
    Copy
  

Learn more about other augmentations…

• Horizontal Flip;
• Vertical Flip;
• Random Crop;
• Random Sized Crop;
• Rotate;
• Resize;
• Blur;
• Center Crop;
• Color Jitter;
• Shift Scale Rotate;
• Longest Max Size;
• Equalize;
• To gray;
• Shear;
• Mosaic;
• Copy Paste.