# Synthetic Examples Improve Generalization for Rare Classes

## TL;DR

This paper demonstrates that augmenting training data with synthetic images significantly improves the classification accuracy of rare classes in image recognition tasks, especially when high variation in simulation is used.

## Contribution

It provides a detailed analysis of how synthetic data can enhance rare class detection and offers best practices for integrating simulated data into training pipelines.

## Key findings

- Synthetic data reduces error rates for rare classes.
- Increasing synthetic data improves accuracy on target classes.
- High variation in simulated data yields maximum performance gains.

## Abstract

The ability to detect and classify rare occurrences in images has important applications - for example, counting rare and endangered species when studying biodiversity, or detecting infrequent traffic scenarios that pose a danger to self-driving cars. Few-shot learning is an open problem: current computer vision systems struggle to categorize objects they have seen only rarely during training, and collecting a sufficient number of training examples of rare events is often challenging and expensive, and sometimes outright impossible. We explore in depth an approach to this problem: complementing the few available training images with ad-hoc simulated data. Our testbed is animal species classification, which has a real-world long-tailed distribution. We analyze the effect of different axes of variation in simulation, such as pose, lighting, model, and simulation method, and we prescribe best practices for efficiently incorporating simulated data for real-world performance gain. Our experiments reveal that synthetic data can considerably reduce error rates for classes that are rare, that as the amount of simulated data is increased, accuracy on the target class improves, and that high variation of simulated data provides maximum performance gain.