Shubham Agrawal

I am a Senior Researcher at Samsung Research America in New York City, where I work on robotics, computer vision, and machine learning. I am fortunate to be working under supervision of Prof. Volkan Isler. Before joining Samsung, I received my CS masters from Columbia University with a focus on robotics and computer vision. During that time, I did several research projects under supervision of Prof. Shuran Song. I did my Bachelors in Computer Science from IIT Kanpur. In past, I also worked at Tesla Inc and Adobe Sytems as software development engineer.

Email  /  CV  /  Google Scholar  /  Github  /  LinkedIn

profile photo
Research

RIC: Rotate-Inpaint-Complete for Generalizable Scene Reconstruction
Isaac Kasahara, Shubham Agrawal, Selim Engin, Nikhil Chavan-Dafle, Shuran Song Volkan Isler
ICRA, 2024
project page / arXiv / code

We present a method for scene reconstruction by structurally breaking the problem into two steps: rendering novel views via inpainting and 2D to 3D scene lifting. Specifically, we leverage the generalization capability of large visual language models (Dalle-2) to inpaint the missing areas of scene color images rendered from different views. Next, we lift these inpainted images to 3D by predicting normals of the inpainted image and solving for the missing depth values.

Real-time Simultaneous Multi-Object 3D Shape Reconstruction, 6DoF Pose Estimation and Dense Grasp Prediction
Shubham Agrawal, Nikhil Chavan-Dafle, Isaac Kasahara, Selim Engin, Jinwook Huh, Volkan Isler
IROS, 2023
project page / arXiv / code

We present a novel method to provide geometric and semantic information of all objects in the scene as well as feasible grasps on those objects simultaneously. The main advantage of our method is its speed as it avoids sequential perception and grasp planning steps

Simultaneous Object Reconstruction and Grasp Prediction using a Camera-centric Object Shell Representation
Nikhil Chavan-Dafle, Sergiy Popovych, Shubham Agrawal, Daniel D. Lee, Volkan Isler
IROS, 2022
video / arXiv / code

We present a new approach to simultaneously reconstruct a mesh and a dense grasp quality map of an object from a depth image. At the core of our approach is a novel cameracentric object representation called the “object shell” which is composed of an observed “entry image” and a predicted “exit image”.

Scene Editing as Teleoperation (SEaT): A Case Study in 6DoF Kit Assembly
Shubham Agrawal*, Yulong Li*, Jen-Shuo Liu, Steven K. Feiner, Shuran Song
IROS, 2022
project page / video / arXiv / code

We create a manipulatable digital-twin of a real-world 6DoF-Kitting scene and propose a deep-learning based approach to infer precise 6DoF object poses inside kit from imprecise poses provided by user.

AdaGrasp: Learning an Adaptive Gripper-Aware Grasping Policy
Zhenjia Xu, Beichun Qi, Shubham Agrawal, Shuran Song
ICRA, 2021
project page / video / arXiv / code

We learn a single grasping policy that generalizes to novel grippers.

Fit2Form: 3D Generative Model for Robot Gripper Form Design
Shubham Agrawal*, Huy Ha*, Shuran Song
CoRL, 2020
project page / video / arXiv / code

Given an object to be grasped, we create a 3D generative model to generate 3D geometry of parallel jaw gripper fingers that optimizes various design objectives.