3D Computer Vision | Active illumination | Multi-view geometry | Generative models
Balachander
Sathianarayanan
About
I work on 3D computer vision under real-world constraints, where standard assumptions—dense views, perfect synchronization, and controlled lighting—often break down. At Columbia, I worked with the CS3 team under Professor Zoran Kostic on traffic scene reconstruction and vision pipelines.
Previously a Research Fellow at RBCDSAI, IIT Madras for 1.5 years, advised by Professor Kaushik Mitra, where I designed and built custom 3D photometric stereo imaging systems for medical applications.
My work lies at the intersection of 3D vision, multi-view geometry, and computational imaging, with a growing interest in extending these systems to 4D scene understanding using generative models.
Research Interests
- 3D Reconstruction & Multi-View Geometry
Understanding and reconstructing visual scenes from multi-view and sparse inputs, including depth estimation, point cloud generation, and scene consistency. - Computational Imaging & Active Illumination
Designing imaging systems that combine optical hardware, structured lighting, and computational methods to recover geometry and surface properties beyond what passive cameras capture. - Geometry-Conditioned Generative Models
Leveraging diffusion models with geometric priors like depth and surface normals for 3D-consistent data augmentation and scene understanding.