Zilverling

G. Pai PhD (Gautam)

Assistant Professor
EEMCS-AM-MIA
Zilverling 2051
g.pai@utwente.nl
+31534896931
Business card

I am an Assistant Professor in the Mathematics of Imaging & AI group (DAMUT-EEMCS) at the University of Twente. My research interests broadly span different aspects of geometry in computer vision and machine learning. I am particularly interested in the applications of spectral and metric geometry for shape analysis, unsupervised learning, and image processing.

Previously, I was a postdoctoral researcher at École Polytechnique - Paris, and at TU/e - Eindhoven and did my PhD at Technion - Haifa

Expertise

  • Mathematics

    • Deep Learning Method
    • Neural Network
    • Convolutional Neural Network
    • Matrix (Mathematics)
    • Numerical Algorithm

  • Engineering

    • Sparse Approximation

  • Computer Science

    • Computer Vision
    • Laplace Operator

Organisations

The guiding theme of my research is the interplay between Geometry, Numerical Algorithms, and Machine Learning. I develop solutions for computational problems from different geometric modalities: images, graphs, and discrete representations of geometric shapes like 3D meshes and point clouds. Recent advances in deep learning provide compelling tools with neural networks to generate state-of-the-art performance for multiple challenges in image and geometry processing. However, despite this progress, they are generally considered to lack mathematical intuition and often demand significant amounts of clean, annotated training data for learning acceptable models. My research aspires to address these shortcomings by establishing a more meaningful merger between mathematically precise algorithms on one end, and data-intensive learning methods on the other. Therefore, besides improving performance, I aspire to improve mathematical interpretability and data efficiency in computational models. You can find some topics of my current interests in an Overview and, my papers with links to all details in Publications

Publications

2025

Loss function inversion for improved crack segmentation in steel bridges using a CNN framework (2025)Automation in construction, 170. Article 105896. Kompanets, A., Duits, R., Pai, G., Leonetti, D. & Snijder, H. H.https://doi.org/10.1016/j.autcon.2024.105896

2024

Geodesic Tracking via New Data-Driven Connections of Cartan Type for Vascular Tree Tracking (2024)Journal of Mathematical Imaging and Vision, 66(2), 198-230. van den Berg, N. J., Smets, B. M. N., Pai, G., Mirebeau, J.-M. & Duits, R.https://doi.org/10.1007/s10851-023-01170-x

2023

Analysis of (sub-)Riemannian PDE-G-CNNs (2023)Journal of Mathematical Imaging and Vision, 65(6), 819-843. Bellaard, G., Bon, D. L. J., Pai, G., Smets, B. M. N. & Duits, R.https://doi.org/10.1007/s10851-023-01147-wFunctional Properties of PDE-Based Group Equivariant Convolutional Neural Networks: 6th International Conference on Geometric Science of Information, GSI 2023 (2023)[Other contribution › Other contribution]. Springer. Pai, G., Bellaard, G., Smets, B. M. N., Duits, R., Nielsen, F. & Barbaresco, F.https://doi.org/10.1007/978-3-031-38271-0_7Geometric Adaptations of PDE-G-CNNs (2023)In Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21–25, 2023, Proceedings (pp. 538-550). Springer. Bellaard, G., Pai, G., Oliván Bescós, J., Duits, R., Calatroni, L., Donatelli, M., Morigi, S., Prato, M. & Santacesaria, M.https://doi.org/10.1007/978-3-031-31975-4_41

2022

Deep Isometric Maps (2022)Image and vision computing, 123. Article 104461. Pai, G., Bronstein, A., Talmon, R. & Kimmel, R.https://doi.org/10.1016/j.imavis.2022.104461Implicit field supervision for robust non-rigid shape matching (2022)In European Conference on Computer Vision (pp. 344-362). Sundararaman, R., Pai, G. & Ovsjanikov, M.

2021

Dpfm: Deep partial functional maps (2021)In 2021 International Conference on 3D Vision (3DV) (pp. 175-185). Attaiki, S., Pai, G. & Ovsjanikov, M.https://doi.org/10.1109/3DV53792.2021.00040Fast sinkhorn filters: Using matrix scaling for non-rigid shape correspondence with functional maps (2021)In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 384-393). Pai, G., Ren, J., Melzi, S., Wonka, P. & Ovsjanikov, M.https://doi.org/10.1109/CVPR46437.2021.00045

2020

On geometric invariants, learning, and recognition of shapes and forms (2020)In Handbook of Variational Methods for Nonlinear Geometric Data (pp. 443-461). Pai, G., Joseph-Rivlin, M., Kimmel, R. & Sochen, N.https://doi.org/10.1007/978-3-030-31351-7_16

Research profiles

Courses academic year 2024/2025

Courses in the current academic year are added at the moment they are finalised in the Osiris system. Therefore it is possible that the list is not yet complete for the whole academic year.

Scan the QR code or
Download vCard