Synthetic data generation for AI: Simmetry.ai secures €330k from NBank

Summary

Simmetry.ai, a DFKI spin-off based in Osnabrück, has secured €330k from NBank to create a platform capable of generating annotated synthetic data for training computer vision models in fields such as agriculture, food, and industrial sectors. The goal is to reduce development time and costs by addressing the traditional data bottleneck through photorealistic simulations and controlled scenarios.

Key takeaways

• The €330k funding from NBank enables Simmetry.ai to accelerate the generation of annotated synthetic data for training AI models, with a focus on agriculture, food, and industries.

• The synthetic data approach aims to overcome the traditional data bottleneck, offering coverage of real-world scenarios that are hard to capture in the real world.

• The platform supports critical tasks like semantic segmentation, object detection, and 3D pose estimation, useful for robotics and industrial inspection applications.

• It is essential to validate synthetic data with real data to avoid simulation bias and ensure robust generalization of AI models.

Simmetry.ai, a German startup based in Osnabrück, announced that it has secured €330k from NBank as part of the High-Tech Incubator (HTI) accelerator. This capital injection accelerates the development of a platform capable of generating annotated synthetic data for AI.

The company was founded in 2024 as a spin-off from the German Research Center for Artificial Intelligence (DFKI) and is led by Kai von Szadkowski (CEO), Anton Elmiger (CTO), and Prof. Dr. Stefan Stiene. The platform enables the generation of controlled datasets that expand coverage of real-world scenarios that are difficult to sample.

The generation of synthetic data does not replace real data, but complements it, offering coverage of rare and costly-to-acquire scenarios.

According to the company, the platform supports activities such as semantic segmentation, object detection, 3D pose estimation and regression, targeting CV engineers and AI developers working in robotics, autonomous machinery, quality inspections, and other contexts in complex and evolving environments. The crucial point is to provide controlled datasets that allow models to tackle edge cases not easily represented by real data.

This type of synthetic data allows testing models in realistic scenarios without having to resort to long-term data collection campaigns.

Simmetry.ai notes that more than 80% of the effort to develop an AI model is devoted to data collection and data preparation. This becomes the key element to address to make AI affordable and scalable.

The company intends to use the funding to develop a scalable platform that enables developers to generate annotated training data for specific use cases, including semantic segmentation and 3D pose estimation, up to regression tasks. The platform is designed to reduce the time and costs needed to build robust models, especially in contexts where real data are scarce or difficult to obtain.

Source eu-startups.com