How real is synthetic data?

Realism check : How ‘real’ are our synthetic images compared to captured images?

Synthetic image data has come a long way in recent years, and at Synthetic Images, we pride ourselves on being a leader in this cutting-edge technology. Our team of experts uses advanced algorithms and machine learning techniques to generate images that are virtually indistinguishable from captured images.

On this page, we’ve assembled a collection of captured vs. synthetic image comparisons that showcase the incredible level of visual realism that we’re able to achieve. Whether it’s the fine details of a surface texture or the play of light and shadow across a scene, you’ll see first-hand just how closely our synthetic images can match the look and feel of the real world.

To increase the realism on a pixel level, we use GANs to transfer the style of your images to the synthetic ones.

Application : Automated visual inspection

Use case : Surface defect detection on metal components.
Demo_RealvsSI_4_real

Captured reference image

Generated synthetic image

Application : Vision guided robotic handling

Use case : Part kitting at an assembly station.
Demo_RealvsSI_3_real

Captured reference image

Generated synthetic image

Application : Automated visual inspection

Use case : Damage detection on pre-cast concrete blocks / bridges.
Demo_RealvsSI_6_real

Captured reference image

Generated synthetic image