Papers
5
Total Citations
319
H-Index
5
About
Abhijit Kundu is a computer vision and robotics researcher whose work spans mobile robot perception, motion segmentation, and 3D object detection. His early research made significant contributions to the challenge of detecting and tracking independently moving objects using monocular cameras mounted on mobile robots — a notoriously difficult problem given the complexity of distinguishing object motion from camera ego-motion. His 2009 paper on multi-view geometric techniques for moving object detection garnered 152 citations, establishing him as a notable voice in dynamic scene understanding. He continued refining these ideas through realtime detection systems and incremental motion segmentation frameworks capable of simultaneously building environmental maps while identifying moving objects — work directly applicable to SLAM-based robotic systems. More recently, Kundu has expanded his focus to autonomous driving perception, particularly 3D object detection in LiDAR point clouds. His innovative application of Long Short-Term Memory (LSTM) networks to incorporate temporal information across LiDAR frames represents a meaningful advance over frame-by-frame detection methods, accumulating nearly 100 citations. Across his career, Kundu has consistently bridged theoretical computer vision with practical robotics applications, demonstrating lasting relevance across both academic and industry-oriented research communities.
Research Focus
Key Achievements
Top Papers
- 1
- 2An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds95 citations · 2020
- 3
- 4Realtime moving object detection from a freely moving monocular camera17 citations · 2010
- 5An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds11 citations · 2020