Trying to understand the world in 3D

I am now at Snap Inc. Previously, I was a postdoctoral researcher at École des Ponts ParisTech and TU Graz. I earned my PhD with distinction at TU Graz under the expert guidance of Professors Vincent Lepetit and Friedrich Fraundorfer. My primary interests revolve around the advancement of Machine Learning and Computer Vision methods within the field of 3D Scene Understanding. I am excited by the possibilities of creating digital twins for indoor environments, with a specific emphasis on reconstructing 3D objects and structural elements in complex scenarios.

📜 Updates

[October 2025] Our GuideFlow3D will be presented at NeurIPS 2025 :book: . We propose guiding pretrained rectified flow model for transferring appearance to 3D assets.

[July 2025] I am now at Snap Inc., pushing limits of augmented reality with Spectacles :sunglasses:

[May 2025] I am honored to be listed among the outstanding reviewers for CVPR 2025. 😁

[May 2025] SCANnotate++ :sparkles: Free high-quality CAD model and pose annotations for the ScanNet++ v1 dataset for you supervised models! Verified by humans :neckbeard: but created automatically :space_invader: with our SCANnotate++ method. To be presented at CVPR 2025 Workshop – Synthetic Data for Computer Vision.

[March 2025] PyTorchGeoNodes (PGN) will be presented at CVPR2025 :boom: We also added some procedural Gaussian Splatting to PGN, check it out!

[August 2024] Our implementation of PyTorchGeoNodes module for parsing shape programs designed in Blender to differentiable PyTorch code is now public. I am aiming to release full algorithms and experiments on ScanNet in early 2025.

[July 2024] I am now in Paris and looking forward to my next research adventures—and, of course, the boulangeries 😁 🥐☕. At École des Ponts ParisTech, I will continue working with Professor Vincent Lepetit and research procedural models \ shape programs \ structured language for applications in 3D scene understanding. I am very grateful to everyone at TU Graz, especially Professor Friedrich Fraundorfer, for helping me grow academically!

[March 2024] 🚀 Our PyTorchGeoNodes is a differentiable module for reconstructing 3D objects from images using interpretable shape programs. I will do my best to release our implementation in August!

[October 2023] 📢 We released our SCANnotateDataset that contains CAD model and pose annotations for objects in the ScanNet dataset.

[October 2023] 📢 Our HOC-Search was accepted at 3DV. Special thanks to my collaborators Stefan Ainetter, Friedrich Fraundorfer and Vincent Lepetit.

[September 2023] I successfully defended my PhD thesis “Playing Proposal Selection Games in 3D Scene Understanding”. Special thanks to my supervisors Prof. Vincent Lepetit (ENPC ParisTech) and Prof. Friedrich Fraundorfer (TU Graz) and to my external reviewer and examiner Prof. Angela Dai (TU Munich). I am looking forward to the next chapter but more updates on that will follow soon! 🤘