Improve Models With The Right Synthetic Data
Generate the right training data for computer vision models and make improvement in weeks instead of months.
Data On Demand
Benchmarked for Quality
Synthetic training data is the fastest and cheapest way to improve or bootstrap a computer vision model. SBX quickly matches real-world sensors and scenes to generate endless simulated variations
SBX synthetic data is generated using technology from film and gaming, to produce realistic, perfectly annotated training datasets for object detection, segmentation, keypoint detection and 6D pose estimation models.
Benchmarked For Quality
Each SBX dataset is the product of iterative testing and optimization to achieve the best performance on real-world data.
Rapid Iteration Support
Request updates to datasets and quickly address new scenes or items as your production data shifts.
Generated Data For Vision
SBX synthetic data is generated using technology from film and gaming, to produce realistic, perfectly annotated training datasets for object detection, segmentation, keypoint detection and 6D pose estimation models.
the right data
for your vision model
Synthetic Data FAQ
Should I use in-house / open source for synthetic data generation?
SBX's mission is to deliver the value of synthetic data as quickly and easily as possible. Creating high quality synthetic data requires a specific set of technology and skills, similar to generating CGI in the VFX industry. SBX was formed around this goal and the primary focus of our team of engineers and technical artists is to get you and your team the best data as efficiently as possible.
Should I start with real training data?
If you are hesitant to begin with synthetic data it makes sense to see how far a small training set of real data can go. SBX can assist with annotation of small batch real data (see hybrid training page)
How much will it cost? How long will it take?
SBX typically aims to create custom synthetic datasets in 10-15 business days. Pricing includes all assets and benchmarking, and so can depend on the type of data being requested. Contact us for more info. SBX is happy to create engagement proposals at no cost.
Will it work for my industry / problem?
SBX has built computer vision training sets for a variety of domains. Check out the industries page to learn more.
What do I need to get started?
Sharing a small amount of real data with SBX is the fastest way to get started. The best validation data to share is both
- diverse : representative of real data
- challenging : rare examples or those with known performance issues on existing vision models.
Do I need to provide 3D models?
No. SBX will create or source 3D assets as needed for a project. If you already have assets for items of interest, we are happy to include and use those in the data SBX produces.
Can SBX match my sensor intrinsics / calibration?
Yes, if you have calibrated your image sensor this information is very helpful to share with SBX. We can match non-linear lens distortion.
What is Synthetic Data?
Synthetic Data is data that’s created artificially based on mathematical algorithms and not real-world events. This makes it an efficient and reliable alternative to standard data.
Benefits of Synthetic Data
- Producing synthetic data takes less time and is highly cost-effective
- Reduces the constraints on obtaining difficult-to-retrieve or tightly regulated data
- Has the ability to be shared and used across industries or with colleagues faster
- Use it to train & pre-train machine learning methods with large data repositories
Bounding Boxes / Masks
2D Keypoints
- 2D rectangle annotations and per-pixel masks for each item of interest in the image.
- Sets of 2D points for specific locations on items of interest in an image
6D Pose
Remove Data as a Blocker and Get Back to Building
Share 25 images from your vision system, and we will generate an optimized training set of 25,000 annotated synthetic images.