Mixing Real and Synthetic Data to Enhance Neural Network Training -- A Review of Current Approaches

TL;DR

This paper reviews various methods that combine real and synthetic data to improve neural network training in computer vision, reducing the need for extensive real-world annotations.

Contribution

It provides a comprehensive comparison of current approaches that enhance neural network training using data augmentation and synthetic data generation.

Findings

Synthetic data can significantly improve model performance.

Annotation-preserving transformations are effective for data augmentation.

Combining real and synthetic data reduces annotation costs.

Abstract

Deep neural networks have gained tremendous importance in many computer vision tasks. However, their power comes at the cost of large amounts of annotated data required for supervised training. In this work we review and compare different techniques available in the literature to improve training results without acquiring additional annotated real-world data. This goal is mostly achieved by applying annotation-preserving transformations to existing data or by synthetically creating more data.