Google AI launches a new dataset generator called Kubric

Researchers Klaus Greff, Google AI researcher François Belletti, Google Scholar Lucas Beyer and several others have published a paper around a scalable dataset generator, Kubric. It is an open source Python framework that uses both PyBullet and Blender to create high quality images. PyBullet is used to train the model to physically simulate interacting with other objects, while Blender is used for rendering images. The tool was designed to reduce the costs and resources associated with generating mature and unbiased real data.

The research paper demonstrated the effectiveness of Kubric using a series of 13 separate data sets that were generated for tasks involved in unsupervised multi-object video detection. The datasets covered several tasks ranging from 3D NeRF models to optical flow estimation. Kubric has released photorealistic scenes that are heavily annotated and can be easily scaled up for larger tasks performed by thousands of machines. The tool is capable of generating huge volumes of such synthetic data.

Even with the urgent need for cheaper, well annotated and unbiased data, there is a lack of software tools that generate efficient and usable data. Synthetic data has become more preferable in the recent past due to its many advantages – cheaper costs, rich annotations, giving researchers full control over their data, and avoiding risks associated with licensing and privacy.

About Mark A. Tomlin

Check Also

Roadmap for a literature review

By Prof. RA Seetha Bandara, Board Member of the Sri …