Resource Data

TrivialAugment: The Next Evolution of Data Augmentation | by Devansh – Machine Learning Simplified | Culture Geek | February 2022

Machine learning (image classification) meets German efficiency

IT costs keep falling. It fueled the tech boom
Our functions can often create augmented images very different from the original. This is a good thing
RandAugment only had 2 hyperparameters but could produce many different images by varying them
RandAugment has outpaced competing data augmentation policies on various architectures
Pseudocode for TA. Unlike RA, the increase force is not fixed.
AutoAugment (AA) and RandAugment (RA) are the most popular augmentation methods. TA manages to outperform
TA’s ability to generalize across different architectures and augmentation spaces is impressive.
An example of one of the many augments available