Agastya Kalra

Papers

Total Citations

137

H-Index

About

Agastya Kalra is a computer vision researcher whose work bridges fundamental perception challenges and real-world industrial deployment. His primary research areas include transparent object segmentation, 6DoF pose estimation, and multimodal sensor fusion for robotics. Kalra’s most impactful contribution is his pioneering use of light polarization cues to solve the notoriously difficult problem of transparent object segmentation—a task where objects lack their own texture and instead adopt background patterns. His 2020 paper on this topic has garnered over 130 citations, establishing a new paradigm for handling optically challenging materials in vision systems. More recently, Kalra has focused on bridging the gap between academic vision research and industrial robotics requirements. His 2024 work introduces the Industrial Plenoptic Dataset (IPD), the first dataset designed for the co-evaluation of cameras, HDR imaging, and algorithms for 6DoF pose estimation. This contribution directly addresses the reliability and accuracy standards needed for mass deployment in manufacturing settings. Through his research, Kalra demonstrates a rare ability to identify fundamental open problems—like transparent object perception—and develop practical, deployable solutions that push the boundaries of what computer vision can achieve in both laboratory and factory environments.