TrivialAugment: Tuning-free Yet State-of-the-Art Data Augmentation

TL;DR

TrivialAugment is a simple, tuning-free data augmentation method that surprisingly outperforms more complex automatic augmentation techniques across various vision tasks, promoting ease of use and reproducibility.

Contribution

We introduce TrivialAugment, a parameter-free augmentation method that simplifies automatic augmentation, achieving state-of-the-art results without extensive tuning or complex procedures.

Findings

TrivialAugment outperforms previous augmentation methods in multiple image classification benchmarks.

It requires no parameter tuning and applies a single augmentation per image.

The method is simple, effective, and easy to adopt in practice.

Abstract

Automatic augmentation methods have recently become a crucial pillar for strong model performance in vision tasks. While existing automatic augmentation methods need to trade off simplicity, cost and performance, we present a most simple baseline, TrivialAugment, that outperforms previous methods for almost free. TrivialAugment is parameter-free and only applies a single augmentation to each image. Thus, TrivialAugment's effectiveness is very unexpected to us and we performed very thorough experiments to study its performance. First, we compare TrivialAugment to previous state-of-the-art methods in a variety of image classification scenarios. Then, we perform multiple ablation studies with different augmentation spaces, augmentation methods and setups to understand the crucial requirements for its performance. Additionally, we provide a simple interface to facilitate the widespread adoption of automatic augmentation methods, as well as our full code base for reproducibility. Since our work reveals a stagnation in many parts of automatic augmentation research, we end with a short proposal of best practices for sustained future progress in automatic augmentation methods.