Synthetic training data is the fastest and cheapest way to improve or bootstrap a computer vision model.
Skip costly hardware setup, data collection, data annotation, and data cleaning. Using technology from film and gaming, we produce realistic, perfectly labeled training datasets for object detection, segmentation, and 6D pose estimation models.
Each SBX dataset is the product of iterative testing and optimization to achieve the best performance on real-world data.
Try it yourself with our free 10K sample dataset & tutorial.
Autonomous forklifts need to keep track of their human-driven counterparts to operate with them side-by-side. The team at OTTO Motors looked to SBX synthetic data to bootstrap and improve computer vision systems in their line of AMRs.
Thanks to a model trained on SBX synthetic data, the ER-FLEX robot can now fetch part bins in a manufacturing facility.
To pick a bin from the shelf, the robot must understand its precise location and orientation (6D pose). Synthetic data offers an easy alternative to the nearly impossible task of collecting this data in the real world.
Our client was building a smart camera to identify table tops in a cafeteria, some of which may still have items on them.
Setting up sensors to collect data across different cafeterias was a major hurdle to scale this client's technology from the first prototype to other locations.
With our synthetic data generator, SBX was able to produce an infinite variety of virtual cafeterias to ensure that our client's vision system adapted to new environments.
Led applied research projects at UberATG, Kindred AI, and SigOpt focusing on computer vision, robotics, and optimization for machine learning.
Built the tech and team behind Wish.com merchant marketplace, growing it to 100k+ active merchants and 45 engineers, PMs, and designers.