Sandika Biswas

About

Sandika Biswas is a researcher advancing the frontier of holistic 3D scene understanding, with a focus on physically plausible human-scene reconstruction from monocular RGB images. Her work addresses a critical challenge in robot perception: generating realistic, physically coherent 3D scenes from a single viewpoint. In her most-cited paper, "Physically Plausible 3D Human-Scene Reconstruction From Monocular RGB Image Using an Adversarial Learning Approach" (2023, 5 citations), she pioneers an adversarial learning framework that overcomes the limitations of traditional optimization-based methods. This approach ensures that reconstructed human-scene interactions are not only geometrically accurate but also physically plausible—a key step toward enabling robots to perceive and interact with dynamic environments. Biswas’s contributions lie at the intersection of computer vision, graphics, and robotics, offering a scalable solution for applications in augmented reality, autonomous navigation, and human-robot collaboration. Her work is notable for its innovative use of adversarial training to enforce physical constraints, setting a new standard for holistic scene reconstruction. With growing citations, Biswas is establishing herself as a rising voice in 3D perception, pushing the boundaries of how machines understand and reconstruct complex, real-world scenes.

Research Focus

Key Achievements

Top Papers

1. 1
   Physically Plausible 3D Human-Scene Reconstruction From Monocular RGB Image Using an Adversarial Learning Approach

   5 citations · 2023

Key Collaborators

Contact & Links