Explore ground truth technology

At understand.ai, we provide cutting-edge ground truth annotation technology to enable you to handle complexity at scale.

Solution Headline One or Two Lines

Introducing our state-of-the-art annotation platform, designed to realize complex ground truth annotation projects. With scalable infrastructure, it effortlessly handles high data volumes and projects of any size. Our platform excels in customized data elevation and workflows, tailored to meet specific project needs. We prioritize compliance, adhering to stringent data privacy and security standards.

The seamless integration of user-friendly tools enables streamlined collaboration between customers and labeling partners. Our automation capabilities significantly reduce manual annotation efforts, making large-scale ADAS/AD programs commercially feasible.

Solution Headline One or Two Lines

Highlight platform features

Multi Sensor

Our ground truth platform excels in multi-sensor-integration, facilitating seamless incorporation and processing of data from multiple lidar sensors. This provides a comprehensive view of complex 3D environments and enables precise annotation.

Multi LiDAR

Intuitive Workflow Editor

The Workflow Editor excels in user-friendliness and customization, providing an intuitive interface to create and modify annotation workflows. It enables streamlined tasks, enhanced efficiency, and offers features like intuitive navigation, task dependencies, and customizable instructions.

Workflow Editor

Workflow steps

Within our Workflow Editor, extensive fine-tuning and customization possibilities of each annotation process step allow for flexible adapting and configuring of individual workflow steps to align with project requirements.

Workflow Editor Steps

Multi-Cloud environment

Our entire system is designed to scale rapidly on multiple cloud platforms to meet volume demands in the most cost-efficient way.

Our automation methodology

How we approach labeling automation


As the industry progresses to a higher level of autonomous driving responsibility shifts from the driver to the manufacturer. Companies developing ADAS/AD systems must increase their validation efforts to ensure system quality and reliability. To sustain large-scale validation projects, extensive labeling automation becomes essential.


We address this challenge with the strategic use of labeling automation. Encompassing initial pre-labeling, attribute definition, and rigorous quality checks, our approach markedly minimizes the need for manual annotation and enables us to precisely annotate large datasets across a wide range of use cases.


By implementing labeling automation, we effectively decouple the linear cost growth of manual labeling work from data volume, enabling us to achieve affordable and consistent high-quality annotations at scale.

Scope of our automation

Our automation capabilities span a wide range of both 2D and 3D automation tasks to successfully meet all requirements for your annotation project.

Scope of automation

The number of iterations makes the difference

Significant automation improvements can be achieved after only 2 training iterations!

Automation Results

Thanks to our labeling automation,
large-scale ground truth annotation projects become feasible

Scope of automation

Reference Case: Automation

Explore this following reference case to gain deeper insight of our successful collaboration with one of our costumers.

Aerial view of a multi-lane highway with vehicles, emphasizing potential for autonomous driving data collection

In this reference case, we successfully delivered 23 million objects as 2D Bounding Boxes as a result of labeling automation. The annotations overachieved set quality targets and were able to be provided within critical time restraints.

The Challenge

  • Validate perception algorithms in a very short amount of time with a high volume of annotations
  • Prove to internal stakeholders and the OEM customer that its perception is working accurately

Key Facts

  • Fully managed annotation service - 2D bounding boxes
  • 27,000 km with an estimated 20 million objects
  • deadline in 12 weeks
  • 4 labeling partners involved across different regions for mitigate COVID impact

Key Achievements

  • Overachieved annotation quality target of 98%
  • Over-delivered on throughout targets (received additional volume from slower supplier)
  • peak throughput 1.4 mio objects per day - accepted annotations
  • understand.ai‘s ML model continuously retrained resulting in 68% decrease in time to accepted annotation*
  • 24/7 operation with daily deliveries