Spatioform Lab

Spatioform (Spatial + Form) Lab: We explore the 3D physical world by integrating machine learning with physical reasoning. Our research focuses on advancing scene understanding, content generation, and intelligent reasoning to develop scalable algorithms that bridge fundamental research and transformative applications across computer vision, machine learning, and robotics.

Scene Understanding

Developing physics-informed methods to recover scene geometry, illumination, and material properties from sparse visual observations. Focus areas include single-image HDR reconstruction, image-based modeling, and novel-view synthesis.

Multimodal ML

Advancing multimodal machine learning models that integrate natural language with visual perception to enable content generation and semantic reasoning across modalities. Focus areas include vision-language model, multimodal representation learning, and generative techniques.

Agentic AI for Physical Reasoning

Developing AI agents that enables complex physical reasoning, adaptive action planning, and robust task execution. Research focuses on multi-agent coordination, zero-shot generalization, iterative learning mechanisms and automation.

Group Members

• Xiaopan Chu — Graduate student, Research (Computer Vision)
• Bowen Shi — Graduate student, Research (Computer Vision)
• Qianyi Li — Graduate student, Research (Multi-Modal ML)
• Lasya Priya Patkam — Graduate student, Research (Robot Learning)
• Jiayan Wang — Graduate student, Research (Robot Learning)
• Tian Wang — Graduate student, Research (LLM)
• Jingtian Zhu — Graduate student, Research (Software Engineering)
• Zhifei Ye — Graduate student, Research (Agentic AI)

Former Members

• Yijie Chen — Master's Project, Beyond 2D: Creating 3D Augmented Reality from a Single RGB Image (Spring 2026)

Acknowledgements