Abhijit Kundu


I currently work at Google Research with the Perception team. My research interest lies in the intersection of the fields of computer vision, machine learning and robotics.

Previouly I finished my PhD in Computer Science from Georgia Tech. At Georgia Tech, I was co-advised by Jim Rehg and Frank Dellaert. Before that, I was at Robotics Research Lab of IIIT Hyderabad, India where I worked with Prof. K Madhava Krishna and Prof. C. V Jawahar.


Apr 1, 2022 :sparkles: Two papers Panoptic Neural Fields and Kubric accepted at CVPR 2022.
May 29, 2021 Some of our work with Google Maps were announced at Google I/O.

selected publications

  1. Panoptic Neural Fields: A Semantic Object-Aware Neural Scene Representation
    Abhijit Kundu, Kyle Genova, Xiaoqi Yin, Alireza Fathi, Caroline Pantofaru, Leonidas Guibas, Andrea Tagliasacchi, Frank Dellaert, Thomas Funkhouser.
    In CVPR 2022
  2. Virtual Multi-view Fusion for 3D Semantic Segmentation
    Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, Caroline Pantofaru.
    In ECCV 2020
  3. 3D-RCNN: Instance-level 3D Object Reconstruction via Render-and-Compare
    In CVPR 2018
  4. Feature Space Optimization for Semantic Video Segmentation
    Abhijit Kundu, Vibhav Vineet, Vladlen Koltun.
    In CVPR 2016
  5. Joint Semantic Segmentation and 3D Reconstruction from Monocular Video
    In ECCV 2014