Build computer vision models 10x faster.

Skip manual labeling.
Produce perfectly annotated datasets in the browser, within hours, and at a fraction of the cost of annotation services or in-house teams.

"After our synthetic data pilot with SBX, we hope never to use real data again!"
- John Novak, Director of Computer Vision

Our synthetic training data is a drop-in replacement for labeled data used to train computer vision systems.

We use video game engines to produce perfectly annotated training datasets for object detection, segmentation, and 6D pose estimation models in common formats like MS COCO. We simulate RGB cameras and RGB+Depth sensors.

What is synthetic training data?

Manual annotation is slow -- 10+ seconds per label

Synthetic data is faster and cheaper.

Setting up hardware and sensors to collect data, training a team of labelers, and pruning labeling errors adds months of R&D and weighs heavily on project budgets.

Once configured, our generator can produce 100,000+ perfectly labeled training images within hours, and scales in the cloud.

Add difficult cases to your datasets.

Whether it is tough lighting conditions, adversarial materials, odd camera angles, or rare SKUs, you can augment your training dataset with difficult training examples and build robust models.

With synthetic data, you have full control of your training dataset.


E-grocery Fulfillment



Agriculture Tech

YCB-Video benchmark dataset

YCB-Video is a large scale dataset developed by researchers at the University of Washington, based on the original YCB work.

We are hosting a free copy of this dataset on S3 to share it with the community and help address known distribution issues.

Download YCB-V (265 GB)
YCB-Video is an academic dataset for 6D object pose estimation

Building towards end-to-end simulation

Vision powered robots will play a massive role in automation in the next decade.

SBX is building towards an end-to-end simulation solution so that these robots can be designed, tested and trained all within our offline simulator.

Founding team

Ian Dewancker, CEO

Led applied research projects at UberATG, Kindred AI, and SigOpt focusing on computer vision, robotics, and optimization for machine learning.

Joshua Kuntz, Engineering

Built the tech and team behind merchant marketplace, growing it to 100k+ active merchants and 45 engineers, PMs, and designers.

Artem Avdacev, Product

Yelp’s anti-fraud expert. Ran a team of 25 engineers & analysts to build machine learning pipelines detecting fraud, spam, and abuse.

Contact Us

Working on a computer vision project in industry or in academia? We want to be your synthetic data partner!
Thanks! We'll be in touch shortly.
Oops! Something went wrong while submitting the form.