Koushik Srivatsan

I am a second year PHD student in the department of Electrical and Computer Engineering at the Johns Hopkins University. I am a member of the VIU Lab advised by Dr. Vishal Patel.

Previously, I was a research assisant affliated with the SPriNT-AI Lab at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), advised by Dr.Karthik Nandakumar and Dr.Muzammal Naseer.

I completed my Bachelors in Electronics and Communication Engineering and Master in Signal Processing from Indian Institute of Information Technology. I completed my master's thesis in the Computer Vision Lab at IIT Madras, advised by Dr. Anurag Mittal.

Email  /  Google Scholar  /  Github  /  Twitter  /  LinkedIn

profile photo

Research Interest

My research focuses on modifying knowledge in large generative models, with an emphasis on both concept erasure for safety and personalization for adaptability. My goal is to ensure that generative models can be responsibly modified and personalized to support trustworthy use in diverse real-world settings.

News

  • [April 2025] - STEREO accepted as a Highlight at CVPR 2025.
  • [February 2025] - One paper accepted to CVPR 2025.
  • [August 2024] - Awarded the JHU-ECE department fellowship for the 2024-2025 academic year.
  • [August 2024] - Joined the Johns Hopkins University as a PHD Student
  • [July 2023] - One paper accepted to ICCV 2023.
  • [July 2023] - One paper was accepted as an ORAL presentation to IJCB 2023.
  • [Feb 2023] - One paper accepted to CVPR 2023.

Research

Representative papers are highlighted.

* denotes joint first authors
STEREO
STEREO: A Two-Stage Framework for Adversarially Robust Concept Erasing from Text-to-Image Diffusion Models
Koushik Srivatsan, Fahad Shamshad, Muzammal Naseer, Vishal M Patel, Karthik Nandakumar
CVPR, 2025 ⭐ Highlight (Top 3% of submissions)
[Code] [Paper]

FLIP-diag FLIP: Cross-domain Face Anti-spoofing with Language Guidance
Koushik Srivatsan, Muzammal Naseer, Karthik Nandakumar
ICCV, 2023
[Project] [Code] [Paper]
FedSIS-diag FedSIS: Federated Split Learning with Intermediate Representation Sampling for Privacy-preserving Generalized Face Presentation Attack Detection
Naif Alkhunaizi*, Koushik Srivatsan*, Faris Almalik*, Ibrahim Almakky, Karthik Nandakumar
IJCB, 2023   (Oral Presentation)
[Code] [Paper]

Face attribute attack Evading Forensic Classifiers with Attribute-Conditioned Adversarial Faces
Fahad Shamshad, Koushik Srivatsan, Karthik Nandakumar
CVPR, 2023
[Project] [Code] [video] [Paper]

Academic Services

  • [September 2024] - Serving as a reviewer for TMLR
  • [June 2024] - Serving as a reviewer for NeurIPS 2024
  • [April 2024] - Serving as a reviewer for ECCV 2024
  • [Feb 2024] - Serving as a reviewer for TPAMI

Source code from Jon Barron's website.