• slide

The platform for synthetic data for training computer vision

Rendered.ai offers a Platform as a Service (PaaS) for data scientists, computer vision engineers, and developers who need to create and deploy unlimited, customized synthetic data generation for machine learning and artificial intelligence workflows, reducing expense, closing gaps, and overcoming bias, security, and privacy issues when compared with the use or acquisition of real-world data.

Challenges with AI
The demand for AI is expanding and data is an essential input for working AI. Organizations focused on physical domains who are pursuing the potential of AI/ML systems inherently rely on data captured using physical sensors and sensor platforms. Unfortunately, real data sets often fail to include rare events or assets, have labeling problems when humans are unable to accurately interpret sensor data, and have security and privacy issues, and collecting real datasets can even incur safety issues when data are physically difficult to collect.

One solution to these issues is simulated or ‘synthetic’ data that has been engineered to behave as if it is real data to AI and that is designed to attain specific outcomes. Gartner has identified that up to 60% of data used to train AI will be synthetic by 2024 with much higher adoption of synthetic data by the end of the decade.

The Rendered.ai solution
In late 2021, Rendered.ai released a commercially available PaaS for synthetic data that enables synthetic data engineers, technical users with modeling and simulation capability, to build and deploy tools for generating physics-based synthetic data to be consumed by data scientists and computer vision (CV) users.  Customers create applications, called channels, that incorporate simulated sensors, sensor platforms, and target content using state-of-the-art 3D modeling technology.  Typically, output datasets are composed of images or video with autogenerated labeling that are used together to train CV algorithms. Channels are used along with hosted job management capability to generate nearly unlimited variety of datasets in a cloud-based, high performance compute environment.

In addition to Rendered.ai content, we can work with 3rd parties to license simulation capabilities for modalities ranging from x-ray to hyperspectral imagery (HSI) and other rendering and simulation tools. We are currently partnering with the Rochester Institute of Technology’s DIRS Laboratory to use their technology for remote sensing customers, for example.

Most early synthetic data adopters focus on expensive, one-off project-based generation of individual data sets. We built a platform because, to be successful with synthetic data, customers require an iterative, engineering workflow in which they can update and adapt data generation to changing needs. We provide a subscription-based business model that enables users to try, iterate, and then integrate synthetic data into their workflows.

Rendered.ai focuses on a single line of business around subscription offerings of our PaaS, coupled with professional services, as needed. Our subscription offerings are available in multiple tiers, primarily differentiated by compute access and collaboration tools, and can be supplemented or extended through enterprise agreements and licensing to 3rd party simulation technologies and content.

We have two standard subscription tiers:
Professional and Enterprise. Enterprise Subscriptions include Enhanced Support Credits that enable customers to access standard tasks and capability such as access to third-party simulators. Subscriptions can be supplemented with professional services when needed."

Featured Resources