I am a computer science researcher primarily focused on 3D computer vision. I work at the Computer Vision Lab in the Dept. of Information Technology and Electrical Engineering at ETH Zürich, where Prof. Dr. Luc Van Gool advises me. As a researcher, I am fascinated by how mathematical construction can precisely represent the perceptual concepts of images, such as 3D scene geometry, motions, lights, and color. I aim to use these mathematical concepts to enable machines for a broader adoption using visual data. Presently, I supervise and work on projects that ETH Zürich Foundation and Google Zurich support. Before ETH Zürich, I was a HDR merit scholarship Ph.D. student at Australian National University (ANU), funded by the Australian Research Council. I received a Ph.D. in Engineering and Computer Science on the thesis titled Non-Rigid Structure from Motion, which was nominated for J. G. Crawford Prize at ANU for Best Ph.D. Thesis 2019.
Research Interests
Geometric Computer Vision, Mathematical Optimization, Abstract Algebra, Geometric AI and Robotics. From time to time, I also investigate topics such as graph theory, topological manifolds and compressed sensing.
Recent News
Recent Awards and Achievements
Current Students and Candidate Collaboration (Masters and Doctorate)
[Topic]: Active SLAM.
[Topic]: Contextual 3D Vision.
[Topic]: Object level active 3D reconstruction.
[Topic]: Large scale 3D data acquisition.
[Topic]: Dense 3D Reconstruction of Indoor Scenes.
[Topic]: Active SLAM.
[Topic]: Online dense 3D reconstruction.
[Topic]: Light physics based 3D reconstruction.
(Interning at Apple Zurich, Summer 22.)
Intern and External Informal Collaborators
[Topic]: Large scale 3D reconstruction.
[Topic]: Neural Architecture Search.
[Topic]: Image based rendering.
[Topic]: Photometric Stereo (concluded with two publication at IEEE/CVF WACV 22).
Advisor, Collaborators and Engineer
Recent Publication

[CVPR 22] Uncertainty Aware Deep Multi-View Photometric Stereo
Authors: Berk Kaya, Suryansh Kumar, Carlos Oliveira, Vittorio Ferrari, Luc Van Gool.Conference on Computer Vision and Pattern Recognition, IEEE, 2022, New Orleans, USA.
Project Website Topic: 3D Computer Vision, Machine Learning

[CVPR 22] Generative Flows with Invertible Attentions
Authors: Rhea Sukthanker, Zhiwu Huang, Suryansh Kumar, Radu Timofte, Luc Van Gool.Conference on Computer Vision and Pattern Recognition, IEEE, 2022, New Orleans, USA.
Project Website Topic: Generative Models, Image Processing

[ICRA 22] A Real-Time Online Learning Framework for Joint 3D Reconstruction and Semantic Seg. of Indoor Scenes
Authors: Davide Menini, Suryansh Kumar, Martin R Oswald, Erik Sandstrom, Cristian Sminchisescu, Luc Van Gool.International Conference on Robotics and Automation, IEEE, 2022, Philadelphia, USA.
Robotics and Automation Letter (RA-L), IEEE, 2022.
Project Website Topic: 3D Computer Vision, Machine Learning.

[WACV 22] Neural Architecture Search for Efficient Uncalibrated Deep Photometric Stereo
Authors: Francesco Sarno, Suryansh Kumar, Berk Kaya, Zhiwu Huang, Vittorio Ferrari, Luc Van Gool.Winter Conference on Applications of Computer Vision, IEEE, 2022, Hawaii, USA.
Project Website Topic: 3D Computer Vision, Machine Learning.

[WACV 22] Neural Radiance Fields Approach to Deep Multi-View Photometric Stereo
Authors: Berk Kaya, Suryansh Kumar, Francesco Sarno, Vittorio Ferrari, Luc Van Gool.Winter Conference on Applications of Computer Vision, IEEE, 2022, Hawaii, USA.
Project Website Topic: 3D Computer Vision, Machine Learning.

[CVPR 21] Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces
Authors: Berk Kaya, Suryansh Kumar, Carlos Oliveira, Vittorio Ferrari, Luc Van Gool.Conference on Computer Vision and Pattern Recognition, IEEE, 2021, Nashville, USA.
Project Website Topic: 3D Computer Vision, Machine Learning

[IJCAI 21] Neural Architecture Search of SPD Manifold Networks
Authors: Rhea Sukthanker, Zhiwu Huang, Suryansh Kumar, Erik Goron. E, Yan Wu, Luc Van Gool.International Joint Conference on Artificial Intelligence, 2021, Montreal, Canada.
Project Website Topic: AutoML, Matrix Manifolds

[T-PAMI 21] Superpixel Soup: Monocular Dense 3D Reconstruction of a Complex Dynamic Scene
Authors: Suryansh Kumar, Yuchao Dai, Hongdong Li.Transactions on Pattern and Machine Intelligence, IEEE, 2021.
Project Website Topic: 3D Computer Vision, Geometry