Open-World 3D Scene Understanding
Class-agnostic and open-vocabulary segmentation of point clouds and 3D scenes.
Research Directions
Compact 3D Scene Representations
Compact, interpretable 3D scene representations using geometric primitives.
Functional & Interactive 3D Scenes
Affordance and functionality reasoning in indoor environments for robotic interaction.
Selected Publications
FunRec: Reconstructing Functional 3D Scenes from Egocentric Interaction Videos
Delitzas, Zhang, Gavryushin, Di Mario, Sun, Dabral, Guibas, Theobalt, Pollefeys, Engelmann, Baráth
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
SuperDec: 3D Scene Decomposition with Superquadric Primitives
Fedele, Sun, Guibas, Pollefeys, Engelmann
IEEE/CVF International Conference on Computer Vision (ICCV), 2025
Oral (top 2.3% of accepted papers)
SpaceControl: Introducing Test-Time Spatial Control to 3D Generative Modeling
Fedele*, Engelmann*, Huang, Litany, Pollefeys, Guibas
International Conference on Learning Representations (ICLR), 2026
OpenFunGraph: Open-Vocabulary Functional 3D Scene Graphs
Zhang*, Delitzas*, Wang, Zhang, Ji, Pollefeys, Engelmann
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
Highlight (top 13% of accepted papers)
SceneFun3D: Fine-Grained Functionality and Affordance Understanding in 3D Scenes
Delitzas, Takmaz, Tombari, Sumner, Pollefeys, Engelmann
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024
Oral (top 3.3% of accepted papers)
OpenMask3D: Open-Vocabulary 3D Instance Segmentation
Takmaz, Fedele, Sumner, Pollefeys, Tombari, Engelmann
Conference on Neural Information Processing Systems (NeurIPS), 2023
Mask3D: Mask Transformer for 3D Semantic Instance Segmentation
Schult, Engelmann, Hermans, Litany, Tang, Leibe
International Conference on Robotics and Automation (ICRA), 2023
1st Place ScanNet Benchmark Challenge
Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes
Chibane, Engelmann, Tran, Pons-Moll
European Conference on Computer Vision (ECCV), 2022
Oral (top 2.7% of accepted papers)
Mix3D: Out-of-Context Data Augmentation for 3D Scenes
Nekrasov, Schult, Litany, Leibe, Engelmann
International Conference on 3D Vision (3DV), 2021
Oral (top 11.36% of accepted papers)
DualConvMesh-Net: Joint Geodesic and Euclidean Convolutions on 3D Meshes
Schult*, Engelmann*, Kontogianni, Leibe
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020
Oral (top 5% of accepted papers)
People
Teaching
Contact
Thinking Shapes Lab
Francis Engelmann
()
Università della Svizzera italiana (USI)
Faculty of Informatics
Via la Santa 1
6962 Lugano
Switzerland
Collaborations
We always welcome highly motivated students and researchers to collaborate on joint projects. Please reach out to , including a brief description of your interests and background (short CV, recent grades).
Prospective Master Thesis students (from USI) should have attended at least one of the lab's courses.