Synthetic images are thus images generated by computer graphics libraries from 3D models. Thanks to the lucrative entertainment industry, computer graphics have indeed come a long way, and rendering engines can nowadays generate highly realistic images from 3D models (such as for video games, 3D animated movies, and special effects). It did not take long for scientists to see the potential for computer vision.
Given some detailed 3D models of the target objects/scenes, it is possible with modern 3D engines to render huge datasets of pseudo-realistic images. With proper scripting, you can, for instance, render images of target objects from every angle, at various distances, with different lighting conditions or backgrounds, and so on. Using various rendering methods, it is even possible to simulate different types of cameras and sensors (for instance, depth sensors such as the Microsoft Kinect or Occipital Structure sensors).
Having full control over the scene...
