What is synthetic data?
Synthetic Images are pictures generated using computer graphics, simulation methods and Artificial Intelligence (AI), to represent reality with high fidelity.
Making use of the vast possibilities that Computer Generated Imagery (CGI), Visual Effects (VFX) and Machine Learning (ML) have to offer, we generate synthetic images that exhaustively represent a 2D / 3D scene. Our process to generate synthetic image datasets is based on ‘Structured Domain Randomization’, whereby each variable in a scene is represented with the specified multitude of values, within the domain relevant parameters.
The datasets we provide contain truly synthetic image data and not augmented versions of real images.
Characteristic factors of our process include the use of CAD information as the basis and the use of CGI and VFX methods combined with
domain specific know-how. This approach also enables us to generate pixel-precise annotations (labels) in parallel to the image generation
process. We offer a variety of labels, as required for the machine vision task at hand.
In industrial machine vision projects, where capturing and annotating real data is resource intensive and not always feasible, synthetic image
datasets we have delivered have proven to be an ideal solution to creating an exhaustive training and validation dataset.
We do not intend to replace real image data – our experience has proven that supplementing a small set of real images with a large set of
synthetic images can provide optimal results in training and validation of machine vision models.
We at Synthetic Images focus on generating and delivering image data within the industrial context – for a variety of use cases and domains.