Sayan Deb Sarkar

I'm a 1st-year PhD student at Stanford University in the Gradient Spaces Group, advised by Prof. Iro Armeni, affiliated with Stanford Vision Lab (SVL).

Before starting PhD, I was a CS master student at ETH Zürich supervised by Prof. Marc Pollefeys, working on aligning real-world 3D environments from multi-modal data. I graduated with a Bachelors in Information Technology from Manipal University, India, where I spent time working on face recognition and medical imaging problems.

In 2020-21, I spent a wonderful time working with Shreyas Hampali, Sinisa Stekovic and Mahdi Rad at Prof. Vincent Lepetit's lab on hand-object pose estimation and monte carlo scene search for 3D scene understanding. I view them as mentors entering research, and strive to learn from them.

I am always looking for research collaborations, get in touch if you have something relevant. If you're around the Bay Area, feel free to reach out for a cup of coffee!

Email  /  CV  /  Google Scholar  /  Github  /  Twitter  /  LinkedIn

• 2024-09 I joined Stanford University for PhD in Computer Vision!
• 2023-08 I will be a research intern at Qualcomm XR Labs, Amsterdam for the Fall!
• 2023-07 SGAligner accepted to ICCV 2023. First first-author submission, first accept!
• 2022-09 I started Computer Science MSc at ETH Zurich!
• 2022-07 Keypoint Transformer accepted as Oral at CVPR 2022!
• 2021-05 I started at Mercedes-Benz R & D as a Computer Vision Research Engineer!
---- show more ----
2021-03 MCSS accepted at CVPR 2021!
2020-06 General 3D Room Layout from a Single View by Render-and-Compare accepted at ECCV 2020!
2020-01 I started as a Research Assistant at ICG, TU Graz with Prof. Vincent Lepetit, supported by a Qualcomm fellowship!

My research interests lie at the intersection of Computer Vision and Machine Learning, specifically in the areas of 3D scene understanding and pose estimation.

We focus on the fundamental problem of aligning pairs of 3D scene graphs whose overlap can range from zero to partial and can contain arbitrary changes. We propose SGAligner, the first method for aligning pairs of 3D scene graphs that is robust to in-the-wild scenarios.

We propose an efficient network architecture for estimating pose of two hands and object during complex interaction. We also release the challenging H2O-3D dataset, which contains two hands interacting with YCB objects.

We propose a Monte-Carlo Tree Search (MCTS) based analysis-by-synthesis method to recover complete scene (3D layout+objects) from a RGB-D scan of the environment.
*Equal contribution

We propose an analysis-by-synthesis method to estimate a 3D layout of the room - walls, floors, ceilings - from a single perspective view. The method recovers complex non-cubiod layouts by solving a constrained discrete optimization problem.

Ray Tracing
Computer Graphics Rendering Competition, Autumn Semester 2022

Implemented a ray tracer with functionalities such as advanced camera models, participating media, photon mapping, Disney BRDF, etc on the Nori framework.