Mathias Bernhard
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Mathias Bernhard has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Benjamin Dillenburger
ETH Zurich
2 shared publications - Anton Savov
ETH Zurich
1 shared publication - Frederick Chando Kim
École Polytechnique Fédérale de Lausanne
1 shared publication - Hong-Bin Yang
École Polytechnique Fédérale de Lausanne (EPFL)
1 shared publication - Hua Chai
California Polytechnic State University, San Luis Obispo, CA, USA
1 shared publication - Ioanna Mitropoulou
ETH Zurich
1 shared publication
Latest Publications
- Open Access
- 1 October 2026
- Published by KEAI Publishing in Frontiers of Architectural Research
- Vol. 15 (5) , 1589-1599
DeepCanvas: Sequential design strategy acquisition in context-aware building footprint synthesis using vision-based deep reinforcement learning
- 1 October 2026
- Published by KEAI Publishing in Frontiers of Architectural Research
- Vol. 15 (5) , 1589-1599
Open AccessDeepCanvas: Sequential design strategy acquisition in context-aware building footprint synthesis using vision-based deep reinforcement learning
- 12 March 2026
- Published by Elsevier in Automation in Construction
- Vol. 185
Design-to-fabrication workflow for 3D concrete printed structures with embedded periodic anticlastic surfaces
- 12 March 2026
- Published by Elsevier in Automation in Construction
- Vol. 185
Design-to-fabrication workflow for 3D concrete printed structures with embedded periodic anticlastic surfaces
- Open Access
- 27 January 2025
- Published by Elsevier in Automation in Construction
- Vol. 171
Curvature-informed paths for shell 3D printing
- 27 January 2025
- Published by Elsevier in Automation in Construction
- Vol. 171
Open AccessCurvature-informed paths for shell 3D printing
- Open Access
- 5 July 2023
- Published by Springer Nature
- p. 53-67
Architectural Sketch to 3D Model: An Experiment on Simple-Form Houses
Transforming sketches into digital 3D models has been an enduring practice in the design process since the first digital turn in architecture. However, 3D modeling is time-consuming, and 3D modeling software usually has a cumbersome interface. Aiming to bridge the gap between sketch and 3D model, we propose a framework that can turn a hand-drawn 2D sketch into a 3D mesh. The user can draw a sketch on the canvas through a web-based interface, and the corresponding 3D model will automatically be generated and shown aside. The 3D model can be downloaded or synchronized into the Rhino directly through Grasshopper.The proposed framework uses a machine learning-based approach to generate a 3D mesh from a single hand-drawn sketch by deforming a template shape. Since the generated models have an uneven surface, we apply an optimization step to refine the form, creating a more usable architectural 3D model with planar faces and sharper edges. We create the Simple House Dataset, which consists of 5000 single-volume houses, to train the neural network. We defined five categories of house typologies - distinguished by roof shape and other geometric properties - and generated 1000 models for each class with parameters chosen randomly. Each model includes a 3D mesh and 20 perspective line drawings from different angles.Although the limitation of the generalization ability makes it unlikely to replace the conventional 3D modeling software today, the fast sketch to 3D transformation allows architects to explore the possibility of various architectural forms and may speed up the design process in the early stage. The code of this project and the Simple House dataset has been published on Github [29].
- 5 July 2023
- Published by Springer Nature
- p. 53-67
Open AccessArchitectural Sketch to 3D Model: An Experiment on Simple-Form Houses
Transforming sketches into digital 3D models has been an enduring practice in the design process since the first digital turn in architecture. However, 3D modeling is time-consuming, and 3D modeling software usually has a cumbersome interface. Aiming to bridge the gap between sketch and 3D model, we propose a framework that can turn a hand-drawn 2D sketch into a 3D mesh. The user can draw a sketch on the canvas through a web-based interface, and the corresponding 3D model will automatically be generated and shown aside. The 3D model can be downloaded or synchronized into the Rhino directly through Grasshopper.The proposed framework uses a machine learning-based approach to generate a 3D mesh from a single hand-drawn sketch by deforming a template shape. Since the generated models have an uneven surface, we apply an optimization step to refine the form, creating a more usable architectural 3D model with planar faces and sharper edges. We create the Simple House Dataset, which consists of 5000 single-volume houses, to train the neural network. We defined five categories of house typologies - distinguished by roof shape and other geometric properties - and generated 1000 models for each class with parameters chosen randomly. Each model includes a 3D mesh and 20 perspective line drawings from different angles.Although the limitation of the generalization ability makes it unlikely to replace the conventional 3D modeling software today, the fast sketch to 3D transformation allows architects to explore the possibility of various architectural forms and may speed up the design process in the early stage. The code of this project and the Simple House dataset has been published on Github [29].