Generating synthetic data for AI trainig:
- annotated images/videos for computer vision training
- 3d scenes for reinforsment learning training (robotics, self driving cars, AGI) to help other AI researchers to train better models.
Because synthetic data is just awesome - it can be unbiased, randomized, cover rare scenarios, look like anything from photorealism to work of art, pre-labeled, confidential and configurable. You can change in any way you want to point your AI to the right direction.
We are using latest research in rendering, physics simulation and AI, expesially computer vision, GANs, NERF, scene understanding to automatically generate 3d models, uv maps, photorealistic textures, physycal propertes, simulation scenarios and scene layouts.