DIPM: Data-driven Inverse Procedural Modelling

• Applicant: Professor Dr.-Ing. Peter Eisert
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 555211318

This project develops new methods for 3D reconstruction of complex objects from monocular or multiview images. These methods exploit learned semantic prior knowledge, represented by procedural models, to provide robust solutions despite partial or noisy data. For this purpose, AI methods for geometry estimation are developed by learning data-driven structural and topological rules and taking them into account as prior knowledge during reconstruction. For this purpose, advances in natural language processing and reinforcement learning are used to make the representation of procedural modeling accessible for reconstruction.

• DFG Programme: Research Grants