Tushar Sharma

Synthetic or Real || Datasets for ML

Synthetic or Real || Datasets for ML

ZEG has reduced the gap between synthetic and real world objects to an extent that synthetic image is completely indistinguishable from a real-world image. This opens whole new possibilities in the Machine Learning domain which involves expensive and time consuming process of capturing data.

Now, you can generate hyper-real datasets at a click of button. Moreover, dataset has in-depth information on 3D scene starting from dimensions, material, camera pose, lighting and almost any desired information. The entire information can be adjusted in real time to suit your machine learning needs. Changing the parameters is as easy as adding a requirement in an excel sheet.

Here is a quick video to show realism as well as parameters:

There are a number of use cases for which our dataset can provide a highly precise and more efficient outcome including:

  • Robotics Vision: Grasping to material identification is now as as easy as 'load (our dataset) and go'
  • Product Defects: Finding product defects would be much easier without any manual work needed to create datasets
  • Classic Segmentation/ Image Search: Dive deeper into classification problems with a whole load of extra information from shape, color to material.

We are excited to provide our datasets as a service to enable AI led companies to improve precision and outcomes.

Please feel free to reach out at

Tushar Sharma