Theodora Kontogianni

Project title

Total3D: Building Universal Representations for Spatial AI

What is your project about?

My project is about developing artificial intelligence that can better understand the physical world around us in 3D. This is important because future AI systems will not only analyse text and images, but also need to understand spaces, objects, and relations in the real world, for example in robotics, AR/VR, and digital twins.  

Today, 3D data exists in many different forms, and it is often processed using separate algorithms. This makes it difficult to transfer knowledge across different datasets, sensors, and applications. In the project, I will develop shared 3D representations that preserve shape, meaning, and spatial relations. Furthermore, the project will investigate how such representations can make it easier for humans to interact with and guide AI systems in the physical world.

How did you become interested in your particular field of research?

My interest in the field grew out of my background as an electrical engineer. I have always been fascinated by how mathematical and technical foundations can be translated into systems that work in the real world. Computer vision became a natural field for me because it connects theory, algorithms, and practical applications: How can a computer learn to understand images, spaces, and objects? Later, I became particularly interested in 3D understanding, because it brings artificial intelligence closer to the physical world that humans and robots actually live in.

What are the scientific challenges and perspectives in your project?

A central challenge is that the same physical world can be described as 3D data in many different ways, depending on the sensor, resolution, and application. A chair, a room, or a building may therefore look very different to an algorithm, even though it is the same object or environment. The project investigates how to develop shared 3D representations that preserve shape, meaning, and spatial relations across such differences. The perspective is to create a more general foundation for spatial AI, which can be used across tasks and make it easier for humans to interact with and guide AI systems in the physical world.

What is your estimate of the impact, which your project may have to society in the long term?

Today, AI systems can work impressively well with text and images. But if we want to move towards more general artificial intelligence, it is not enough for AI to read and see. It must also understand the physical world in 3D: where things are, how they relate to each other, and how one can act within a space. In the long term, my research can help give AI this spatial understanding. The goal is to develop methods that can bring different types of 3D data into a shared understanding, so systems can learn across sensors, environments, and tasks. This is an important step towards autonomous agents and robots that can act meaningfully in complex environments.

Which impact do you expect the Sapere Aude programme will have on your career as a researcher?

It is a great honor to receive a Sapere Aude grant. The programme will allow me to take a decisive step as an independent research leader. With the grant, I can build a young research team around a long-term vision: developing spatial AI that brings artificial intelligence closer to the physical world. I look forward to continuing to grow as a researcher and research leader and to strengthening my international network. In the longer term, the programme will contribute to establishing DTU as an international hub for research in 3D understanding, spatial AI, and autonomous agents, and create a strong foundation for future Danish and European research grants.