Aniket Roy

I am a Ph.D. student at Johns Hopkins University in the department of Computer Science under the guidance of Prof. Rama Chellappa. I am currently working on Computer Vision and Machine Learning, specifically on a few-shot learning, multimodal learning, generative AI, including diffusion models and large language models. Prior to that, I obtained M.S.(by Research) degree from Dept. of Computer Science, Indian Institute of Technology Kharagpur. I am honored to be awarded as an Amazon Fellow.

Email  /  CV  /  Google Scholar  /  Twitter  /  Github

profile photo
News

  • 12/2023: Presented Invited talk on Advances in few-shot learning at Indian Statistical Institute Kolkata
  • 10/2023: Awarded as an Amazon Fellow as a part of JHU + Amazon initiative for Interactive AI!
  • 10/2023: Certified Margin Maximization got accepted in NeurIPS'23
  • 09/2023: Halp got accepted in CVPR'23
  • 06/2023: Started Research Internship at SRI International
  • 08/2022: Awarded Murkos Thomas Memorial Award for Best Research Paper in CSE from IIT Kharagpur.
  • 08/2022: FeLMi got accepted in NeurIPS'22
  • 07/2022: MuLOT got accepted in WACV'22
  • 06/2022: Started Research Internship at Amazon AWS AI
  • 06/2021: PASS got accepted in ICCV'21
  • 06/2021: Started as a Research Intern at MERL

Research

My broad research interest lies in the intersection of Machine Learning and Computer Vision. Specifically, I have worked on few-shot learning, multi-modal learning, generative models, foundational models, and Large Language models.

STEPs DiffNat: Improving diffusion image quality using natual image statistics
Aniket Roy, Maitreya Suin, Anshul Shah, Ketul Shah, Jiang Liu, Rama Chellappa
Submitted

We propose a kurtosis loss to generate better quality images

STEPs Cap2Aug: Caption guided Image-to-Image data Augmentation
Aniket Roy, Anshul Shah, Ketul Shah, Anirban Roy, Rama Chellappa
Submitted

We use pretrained caption and diffusion model to generate semantic augmentations.

FeLMI FeLMi : Few shot Learning with hard Mixup
Aniket Roy, Anshul Shah, Ketul Shah, Prithviraj Dhar, Anoop Cherian, Rama Chellappa
NeurIPS 2022

We propose hard mixup to improve few shot learning

HaLP HaLP: Hallucinating Latent Positives for Skeleton-based Self-Supervised Learning of Actions
Anshul Shah, Aniket Roy*, Ketul Shah*, Shlok Mishra, David Jacobs, Anoop Cherian, Rama Chellappa
CVPR 2023

We hallucinate latent postives for learning skeleton encoders without labels

DiffAlign DiffAlign : Few-shot learning using diffusion based synthesis and alignment
Aniket Roy, Anshul Shah*, Ketul Shah*, Anirban Roy, Rama Chellappa
Under Submission

We leverage the recent success of the generative models for few-shot learning

arXiv
DiffAlign Multimodal Learning using Optimal Transport for Sarcasm and Humor Detection
Aniket Roy*, Shraman Pramanick*, Vishal Patel
WACV 2022

We propose optimal transport for multimodal fusion

DiffAlign PASS: Protected Attribute Suppression System for Mitigating Bias in Face Recognition
Prithviraj Dhar*, Josh Gleason*, Aniket Roy, Carlos Castillo, Rama Chellappa
ICCV 2021

We propose an adversarial framework for gender and racial bias mitigation

DiffAlign Distill and De-bias: Mitigating Bias in Face Recognition using Knowledge Distillation
Prithviraj Dhar, Josh Gleason, Aniket Roy, Carlos Castillo, P.J. Philips, Rama Chellappa
arxiv

We propose a knowledge distillation based framework for gender and racial bias mitigation

DiffAlign Digital Image Forensics - Theory and Implementation
Aniket Roy, Rahul Dixit, Ruchira Naskar and Rajat Subhra Chakraborty
Springer, ISBN: 978-9811076435

The book covers topics in digital image forensics

DiffAlign Towards Optimal Prediction Error Expansion based Reversible Image Watermarking
Aniket Roy, Rajat Subhra Chakraborty
IEEE Transactions on Circuits and Systems for Video Technology (IEEE TCSVT)

We performed theoretical analysis of Prediction Error Expansion based Reversible Image Watermarking

DiffAlign Classification of Computer Generated and Natural Images based on Efficient Deep Convolutional Recurrent Attention Model
Diangrati Tariang, Prithviraj Sengupta, Aniket Roy, Rajat Subhra Chakraborty and Ruchira Naskar
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2019

We propose a deep convolutional recurrent attention model for classifying computer generated and natural images.

DiffAlign Discrete Cosine Transform Residual Feature Based Filtering Forgery and Splicing Detection in JPEG Images
Aniket Roy, Diangrati Tariang, Rajat Subhra Chakraborty and Ruchira Naskar
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2018

We propose to use Discrete Cosine Transform Residual Feature Based Filtering Forgery and Splicing Detection

DiffAlign Camera Source Identification Using Discrete Cosine Transform Residue Features and Ensemble Classifier
Aniket Roy, Rajat Subhra Chakraborty, Udaya Sameer, Ruchira Naskar
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2017

We propose to use Discrete Cosine Transform Residual Feature for Camera Source Identification

DiffAlign Copy Move Forgery Detection with Similar But Genuine Objects
Aniket Roy, Akhil Konda and Rajat Subhra Chakraborty
IEEE International Conference on Image Processing (ICIP) 2017

We use Rotated Local Binary Pattern features for similar but genuine object based forgery

DiffAlign Automated JPEG Forgery Detection with Correlation based Localization
Diangarti Tariang, Aniket Roy, Rajat Subhra Chakraborty and Ruchira Naskar
IEEE International Conference on Multimedia and Expo (ICMEW) Workshops 2017

We detect and localize JPEG image forgery based on noise correlation

DiffAlign [Best Paper Award] Optimal Distortion Estimation For Prediction Error Expansion Based Reversible Watermarking
Aniket Roy and Rajat Subhra Chakraborty
International Workshop on Digital-Forensics and Watermarking (IWDW), 2016

We performed theoretical analysis of Prediction Error Expansion Based Reversible Watermarking

DiffAlign An HVS Inspired Robust Non-blind Watermarking Scheme in YCbCr Space Additionally Designed to Prevent Singular Value Exchange Attack
Aniket Roy, Arpan Kumar Maiti and Kuntal Ghosh
International Journal of Image and Graphics (IJIG)

We proposed a human visual system inspired watermarking scheme and performed robustness analysis

DiffAlign Reversible Color Image Watermarking in the YCoCg-R Color Space
Aniket Roy, Rajat Subhra Chakraborty and Ruchira Naskar
ICISS 2015

We propose a reversible color image watermarking scheme and perform information theoretic analysis

Awards and Achievements

  • Awarded as an Amazon Fellow as a part of JHU + Amazon initiative for Interactive AI!
  • Awarded Murkos Thomas Memorial Award for Best Research Paper in CSE from IIT Kharagpur.
  • Recipient of Best Paper Award in International Workshop on Digital-forensics and Watermarking (IWDW)
  • Awarded IEEE Signal Processing Society Travel Grant for attending ICIP 2017
  • Awarded Microsoft Travel Grant for attending CVPR 17 and IWDW 2016.
  • Awarded with SCHEME OF SCHOLARSHIP FOR COLLEGE AND UNIVERSITY STUDENTS during Bachelor's degree course.



Template Credits : Jon Barron