The effects of using created synthetic images in computer vision training

TL;DR

This study demonstrates that synthetic images generated from rendering engines like Unreal Engine 4 can effectively supplement real datasets in computer vision training, reducing the need for large amounts of real data while maintaining high accuracy.

Contribution

It provides a practical methodology for creating and evaluating synthetic images to enhance deep learning model training in data-scarce scenarios.

Findings

Adding over 60% synthetic images narrows accuracy gap to 1-2%.

Synthetic images can replace up to 90% of real data during training.

Using synthetic images reduces data collection costs significantly.

Abstract

This paper investigates how rendering engines, like Unreal Engine 4 (UE), can be used to create synthetic images to supplement datasets for deep computer vision (CV) models in image abundant and image limited use cases. Using rendered synthetic images from UE can provide developers and businesses with a method of accessing nearly unlimited, reproducible, agile, and cheap training sets for their customers and applications without the threat of poisoned images from the internet or the cost of collecting them. The validity of these generated images are examined by testing the change in model test accuracy in two different sized CV models across two binary classification cases (Cat vs Dog and Weld Defect Detection). In addition, this paper provides an implementation of how to measure the quality of synthetic images by using pre-trained CV models as auditors. Results imply that for large (VGG16) and small (MobileNetV3-small) parameter deep CV models, adding >60% additional synthetic images to a real image dataset during model training can narrow the test-training accuracy gap to ~1-2% without a conclusive effect on test accuracy compared to using real world images alone. Likewise, adding <10% additional real training images to synthetic only training sets decreased the classification error rate in half, then decreasing further when adding more real training images. For these cases tested, using synthetic images from rendering engines allow researchers to only use 10% of their real images during training, compared to the traditional 50-70%. This research serves as an example of how to create synthetic images, guidelines on how to use the images, potential restrictions and possible performance improvements for data-scarce projects.