SBX creates synthetic training data for computer vision models in robotics — letting you train better models at lower cost in less time.
The advantages of synthetic training data are well known, especially in real-world settings like robotics. You can get lots of training data quickly, rather than waiting for robots to generate it "manually". You can over-train on important edge cases that are rare in the real world, but as common as you want in sim. Ground truth labels are laborious in the real world, but easy in sim. Etc.
Historically, DL teams have had to create their own training simulation regimes, even though there's a lot of overlap between different projects. SBX solves this problem by specializing in creating high-quality, easily-used synthetic training data for roboticists.