Indu Panigrahi

Hi! I'm a Ph.D. student in the Siebel School for Computing and Data Science at the University of Illinois Urbana-Champaign where I am advised by Prof. Tal August.

My research interests generally lie in human-computer interaction, machine learning, and explainable AI. During my Master's, I primarily focused on explainability in computer vision from a human-centered perspective, and during my undergrad, I worked on developing data-efficient computer vision techniques for analyzing a rock sample dataset.

I previously completed my Master's and Bachelor's in the Department of Computer Science at Princeton where I worked with Prof. Parastoo Abtahi, Dr. Ruth Fong, and Prof. Adam Maloof.

Email  /  Google Scholar  /  GitHub  /  LinkedIn  /  BlueSky

profile photo
Publications & Preprints
Image Interactivity x Explainability: Toward Understanding How Interactivity Can Improve Computer Vision Explanations
Indu Panigrahi, Sunnie S. Y. Kim*, Amna Liaqat*, Rohan Jinturkar, Olga Russakovsky, Ruth Fong, Parastoo Abtahi
CHI Late-Breaking Work, 2025
extended abstract / project page / CHI talk / Master's thesis talk

Explored how end-users leverage and perceive interactive computer vision explanations.
Image Improving Data-Efficient Fossil Segmentation via Model Editing
Indu Panigrahi, Ryan Manzuk, Ruth Fong, Adam Maloof
CVPR Workshop on Learning with Limited Labelled Data for Image and Video Understanding, 2023
Also presented at CVPR 2023 Women in Computer Vision Workshop
Computer Science Outstanding Independent Work Award
paper / poster / 5 min. talk / 8 min. talk

Extended a model-editing technique and conducted extensive experiments to investigate and mitigate systematic mistakes made by a model for fossil segmentation. Originally for my Spring 2022 Junior Independent Work.
Image Leveraging Structure from Motion to Localize Inaccessible Bus Stops
Indu Panigrahi, Tom Bu, Christoph Mertz
arXiv, 2022
Work done while interning at the Robotics Institute as a RISS scholar
preprint / poster / talk / code

Developed a data-efficient computer vision method to detect hazardous, snow-covered sidewalks in images from bus routes by combining Structure from Motion with a segmentation model.
Image Multispectral Petrography Image Analysis as a Method for Paleoenvironmental and Paleoecological Reconstruction; Applications to Earth's First Reefs
Ryan Manzuk, Bolton Howes, Emily Geyman, Indu Panigrahi, Devdigvijay Singh, Adam Maloof
AGU Fall Meeting, 2022
abstract

This presentation covered work in the Maloof Research Group on using petrographic imaging and computer vision to bring high-throughput, fine-scale, geophysical data to the study of carbonate outcrops and Earth history.
More Research Experiences

In addition to the experiences that led to the publications/preprints above, I've had the opportunity for other (related) research experiences!

Image Is it Cake? A Recipe for Data-Efficient Fossil Identification
Indu Panigrahi (Advised by Dr. Ruth Fong & Prof. Adam Maloof)
Senior Thesis, 2022-2023
Outstanding Computer Science Senior Thesis Prize
thesis / talk

Created a data-efficient and generalizable computer vision model that leverages self-supervised learning and curriculum learning to improve the segmentation of fossils from an impactful rock sample dataset.
Image
Incorporating Inter-Image Communication to Improve Serial Image Segmentation
Indu Panigrahi (Advisor: Prof. Adam Maloof, Secondary Advisor: Prof. Jia Deng)
Junior Independent Work, Fall 2021
report / talk / code

Designed and implemented an inter-image communication mechanism that manipulates the Region Proposal Network of a Mask R-CNN to improve the segmentation consistency of a serial image dataset.
Image
Automating Three-dimensional Modeling of an Archaeocyathid reef using a Mask R-CNN
Indu Panigrahi (Advised by Dr. Ryan Manzuk & Prof. Adam Maloof)
ReMatch+ Summer Program, Summer 2021
talk

Using a Mask R-CNN, we segmented an image stack of cross sections of a rock sample encasing extinct reef-building organisms and stacked the segmented images to form a three-dimensional model of the embedded specimens.
Course Projects
Image Exploring Tokenization Techniques for Protein Language Models
Creston Brooks* and Indu Panigrahi*
AI for Precision Health (COS 557), Spring 2024
paper / code

We compared different tokenization techniques for protein sequences to combine amino acid-level information with sequence-level motifs.

Image Comparing Importance Sampling Based Methods for Mitigating the Effect of Class Imbalance
Indu Panigrahi* and Richard Zhu*
Fundamentals of Deep Learning (COS 514), Fall 2023
arXiv / code

Using Planet's Amazon Rainforest dataset, we investigate and compare the effectiveness of three techniques for mitigating data imbalance that derive from importance sampling: loss reweighting, undersampling, and oversampling.

Image Smibbit: The World’s First Public Wash Trading Detector
Indu Panigrahi* and Alexis Sursock*
Elements of Decentralized Finance (ECE 473), Spring 2023
paper / poster

We implemented several methods for detecting wash trading and built a user interface that utilized these methods to flag user-specified NFTs.
DeCenter Spring Conference Outstanding Poster Prize

Image Implementing a vision-controlled quadrotor
Albert Lin*, Raymond Liu*, Indu Panigrahi*, Nobline Yoo*
Introduction to Robotics (MAE 345), Fall 2022

We worked in a team to implement a vision-based controller on a quadrotor using concepts from motion planning, control, localization, and computer vision.

Image OSCAR: Occluding Spatials, Category, And Region under discussion
Raymond Liu* and Indu Panigrahi*
Natural Language Processing Final Project (COS 484), Spring 2022
poster / paper / code

Reproduced a Visual Dialog model from an EMNLP 2021 paper and performed novel ablations and experiments to analyze and improve the input embedding.

Image
Image
Pedestrian Detection and Interpretability
Raymond Liu* and Indu Panigrahi*
Computer Vision Final Project (COS 429), Fall 2021
poster / paper / code

We enhanced pedestrian detection with Faster R-CNN by modifying the loss function to upweight images that lack visual cues, such as crosswalks.

Image
Image
TigerTools
Indu Panigrahi (lead)*, Raymond Liu*, and Adam Rebei*
COS 333 Term Project, Spring 2021
documentation / code

This application allows Princeton University users to find amenities using a map of campus and provide feedback on those amenities. By Indu Panigrahi '23, Raymond Liu '23, and Adam Rebei '23.
Now hosted by TigerApps.

Teaching & Outreach
Image
Department of Computer Science, Princeton University
  • Graduate TA: COS 217 Introduction to Programming Systems (Spring 2025, Fall 2024, Spring 2024, Fall 2023)
  • Undergraduate TA: COS 302 Mathematics for Numerical Computing and Machine Learning (Fall 2022, Spring 2022)
  • Undergraduate Grader: COS 217 Introduction to Programming Systems (Spring 2023, Fall 2021, Spring 2021)
Outstanding Student Teaching Award (Undergrad)
Image
Office of Undergraduate Research, Princeton University

As a ReMatch+ alumna, I often volunteer for OUR as a

  • Mentor for the ReMatch+ and OURSIP programs: Hosted weekly support sessions for summer interns at Princeton to discuss directions for progressing their research and ideas for making their concluding presentations accessible to a general audience. (Summer 2023, Summer 2024)
  • Judge for Princeton Research Day: Evaluated video submissions based on clarity and accessibility to non-expert audiences. (2023, 2024, 2025)
  • Student Outreach Volunteer: Help with ad-hoc information sessions and outreach events.
Image
AI4ALL Summer Camp, Princeton University
  • Volunteer Research Instructor: Taught and helped develop an AI/ML curriculum for high school students underrepresented in AI research. (Summer 2023, Summer 2024)

    In 2023, my co-instructors and I designed and mentored a project on exploring the effect of data on ML models by applying computer vision to satellite images to track deforestation (Media Coverage). In 2024, I played a supporting instructor role and helped guide students through the results and final presentation portion of a project that tackled data imbalance in medical imaging.
Image
RoboLaunch, Carnegie Mellon University
  • Workshop Co-organizer: Co-organized an introductory workshop on PID control for a robotics outreach initiative while interning at CMU. (Summer 2022, Recording)
Committee Service