Skip to main content Skip to secondary navigation

Gradient Spaces Lab

Main content start

There was a truck in the image. We removed it with AI. Can you spot where it was?
Photo taken in October 2024, Stanford University. Photo Credits: Geoffrey Tuttle

About Us

Welcome to the Gradient 𝚫 Spaces Research Group. The group belongs to the Civil and Environmental Engineering Department, Stanford University, under the Schools of Engineering and Sustainability. Our research and educational activities focus on developing quantitative and data-driven methods that learn from real-world visual data to design and construct data-driven sustainable and adaptive environments across the physical and digital space. Of particular interest is the creation of spaces that blend from the 100% physical (real reality) to the 100% digital (virtual reality) and anything in between, with the use of mixed reality and multi-level design (i.e., of buildings, processes, UXs, etc.). We believe that by cross-pollinating the two domains, we can achieve higher immersion and view these spaces as a step toward more equitable living conditions. Hence, we aim for developing methods that work in real-world settings on a global scale. To achieve the above, we are building a cross- and inter- disciplinary team that is diverse and well-rounded. Most importantly, we are driven by curiosity and learning, and so does everything we do.

To learn more about this vision, you can read this short story to illustrate this future and the impact on designers: 
A Day in the Life of an Architect in the Gradient World

News

Research updates

Two papers accepted to NeurIPS 2025, see you in San Diego!

Curious which?

Research updates

One paper accepted to ICCV 2025, see you in Hawaii!

Curious which?

Fellowships & Awards (AY 2024-5)

In the past academic year, here are the awards that our group members got:

 

Research Highlights

GuideFlow3D: Optimization-Guided Rectified Flow For Appearance Transfer
Sayan Deb Sarkar, Sinisa Stekovic, Vincent Lepetit, Iro Armeni
NeurIPS 2025
[pdf] [website] [Code]

SGAligner++: Cross-Modal Language-Aided 3D Scene Graph Alignment
Binod Singh*, Sayan Deb Sarkar*, Iro Armeni
Arixv preprint
[pdf] [website

Rectified Point Flow: Generic Point Cloud Pose Estimation
Tao Sun*, Liyuan Zhu*, Shengyu Huang, Shuran Song, Iro Armeni
NeurIPS 2025 [Spotlight]
[pdf] [website] [code]