Kaustubh Sadekar

I am a Computer Science Ph.D. student at Portland State University, advised by Prof. Atul Ingle. My research interests include 3D computer vision, computational Imaging, and Design of Synthetic dataset generation pipelines.

Previously, I was a Senior Research Fellow at the Computer Vision Graphics and Imaging Lab of IIT Gandhinagar, supervised by Prof. Shanmuganathan Raman. Before that I worked at the Autonomous Robots and Multi-robot Systems (ARMS) Lab of the Indian Institute of Technology Bombay as a Research Associate supervised by Prof. Leena Vachhani and Prof. Abhishek Gupta.

Email  /  CV  /  Google Scholar  /  GitHub  /  LinkedIn  /  YoutTube

Publications
Single-Photon 3D Imaging with Equi-Depth Photon Histograms
Kaustubh Sadekar, David Maier, Atul Ingle,
ECCV, 2024 project page/ arXiv/ video/ Code

Our work paves the way towards lower bandwidth and reduced in-pixel memory requirements for single-photon 3D cameras, making them attractive for resource-constrained 3D vision applications.

3D sensing with single-photon cameras for resource-constrained applications
Kaustubh Sadekar, David Maier, Atul Ingle,
CVPR (CCD Workshop), 2024
poster/ video

We propose a new 3D imaging technique to reduce the bandwidth and in-pixel memory requirements for single-photon 3D cameras.

Shadow Art Revisited: A Differentiable Rendering Based Approach
Kaustubh Sadekar, Ashish Tiwari, Shanmuganathan Raman,
WACV, 2022
project page/ arXiv/ video

We use differentiable rendering to create artistic 3D sculptures for shadow art effects.

LS-HDIB: A Large Scale Handwritten Document Image Binarization Dataset
Kaustubh Sadekar, Ashish Tiwari, Prajwal Singh, Shanmuganathan Raman,
ICPR, 2022
project page/ arXiv/ video

We propose a large scale dataset of 1 Million challenging images for the task of hand written document image binarisation (HDIB) with accurate segmentation groundtruth.

TreeGCN-ED: Encoding Point Cloud using a Tree-Structured Graph Network
Prajwal Singh, Kaustubh Sadekar, Shanmuganathan Raman,
arXiv, 2021
project page/ arXiv/ video

We propose a large scale dataset of 1 Million images for the task of hand written document image binarisation with accurate segmentation groundtruth.

Projects
Lensless Imaging: Replacing a Camera Lens with Scotch Tape!
project page

Development of a low-cost lensless camera and its calibration.

Simulating Single Photon Cameras for Supervised Depth Estimation
project page

Physics-based modeling of SPAD sensor measurements simulating the effect of photon randomness, laser characteristics, and background photons.

Affordable Stereo Camera
project page

This project aims to create a custom RGB-D stereo camera using low cost USB cameras and providing library for calibration, stereo depth estimation and RGB-D to 3D pointcloud reconstruciton.

Implementation of Kinect Fusion Algorithm For 3D Reconstruction
project page

A Python implementation of the Kinect Fusion algorithm with PyTorch based vectorization to enable GPU acceleration capabilities.

OmniCV-Lib - Library for omnidirectional cameras
project page

Virtual camera classes in Python and C++ for various omnidirectional/fisheye/wide-lens camera models with ROS support.

Cyclops - A Spatial AI based Assistant for Visually Impaired People
Kaustubh Sadekar, Malav Bateriwala, Vishruth Kumar
Project Page / Competetion Results

A OAK-D depth camera based system that guides the user to target a object using object detection and depth estimation.

OpenCV Spatial AI Competetion Phase 1 Winners

FUNdamentals of CV

A series of fun projects explaining classical computer vision and image processing topics like calibration, camera projection, color detection etc.

Interactive 3D window using Numpy and OpenCV

Python and C++ implementation for perspective camera projection using OpenCV and Numpy to create an interactive 3D window effect. Also implimented a Kalman filter for smooth face traciking.

SPARK - OpenCV and ROS based robotic sketching arm

A low cost 3DOF robotic sketching arm was fabricated and developed using inexpensive servo motors. A visual feedback system, utilizing ArUco markers, was designed for efficient performance, and closed loop position control of the arm joints was implemented.

Fun mirrors using OpenCV

Created a virtual camera model and used mesh based warping to generate sum fun effects.

Invisibility cloak using OpenCV

Used OpenCV for color based segmentation to create a fun application of invisibility cloak.

Air drums using OpenCV

Used color based segmentation to create a fun application to play drums in air!


Thanks to Jon Barron for this awesome template and Pratul Srinivasan for additional formatting.
Last updated July 2022.